1 /*
2 * linux/fs/ext4/super.c
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>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
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>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
43
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
46
47 #include "ext4.h"
48 #include "ext4_extents.h"
49 #include "ext4_jbd2.h"
50 #include "xattr.h"
51 #include "acl.h"
52 #include "mballoc.h"
53
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
56
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
62
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
66 static int ext4_commit_super(struct super_block *sb, int sync);
67 static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69 static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_sync_fs(struct super_block *sb, int wait);
72 static const char *ext4_decode_error(struct super_block *sb, int errno,
73 char nbuf[16]);
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static void ext4_write_super(struct super_block *sb);
78 static int ext4_freeze(struct super_block *sb);
79 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
80 const char *dev_name, void *data);
81 static inline int ext2_feature_set_ok(struct super_block *sb);
82 static inline int ext3_feature_set_ok(struct super_block *sb);
83 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
84 static void ext4_destroy_lazyinit_thread(void);
85 static void ext4_unregister_li_request(struct super_block *sb);
86 static void ext4_clear_request_list(void);
87
88 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
89 static 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 };
96 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97 #else
98 #define IS_EXT2_SB(sb) (0)
99 #endif
100
101
102 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103 static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
105 .name = "ext3",
106 .mount = ext4_mount,
107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
109 };
110 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
111 #else
112 #define IS_EXT3_SB(sb) (0)
113 #endif
114
ext4_kvmalloc(size_t size,gfp_t flags)115 void *ext4_kvmalloc(size_t size, gfp_t flags)
116 {
117 void *ret;
118
119 ret = kmalloc(size, flags);
120 if (!ret)
121 ret = __vmalloc(size, flags, PAGE_KERNEL);
122 return ret;
123 }
124
ext4_kvzalloc(size_t size,gfp_t flags)125 void *ext4_kvzalloc(size_t size, gfp_t flags)
126 {
127 void *ret;
128
129 ret = kzalloc(size, flags);
130 if (!ret)
131 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
132 return ret;
133 }
134
ext4_kvfree(void * ptr)135 void ext4_kvfree(void *ptr)
136 {
137 if (is_vmalloc_addr(ptr))
138 vfree(ptr);
139 else
140 kfree(ptr);
141
142 }
143
ext4_block_bitmap(struct super_block * sb,struct ext4_group_desc * bg)144 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
145 struct ext4_group_desc *bg)
146 {
147 return le32_to_cpu(bg->bg_block_bitmap_lo) |
148 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
149 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
150 }
151
ext4_inode_bitmap(struct super_block * sb,struct ext4_group_desc * bg)152 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
153 struct ext4_group_desc *bg)
154 {
155 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
156 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
157 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
158 }
159
ext4_inode_table(struct super_block * sb,struct ext4_group_desc * bg)160 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
161 struct ext4_group_desc *bg)
162 {
163 return le32_to_cpu(bg->bg_inode_table_lo) |
164 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
165 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
166 }
167
ext4_free_group_clusters(struct super_block * sb,struct ext4_group_desc * bg)168 __u32 ext4_free_group_clusters(struct super_block *sb,
169 struct ext4_group_desc *bg)
170 {
171 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
172 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
173 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
174 }
175
ext4_free_inodes_count(struct super_block * sb,struct ext4_group_desc * bg)176 __u32 ext4_free_inodes_count(struct super_block *sb,
177 struct ext4_group_desc *bg)
178 {
179 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
180 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
181 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
182 }
183
ext4_used_dirs_count(struct super_block * sb,struct ext4_group_desc * bg)184 __u32 ext4_used_dirs_count(struct super_block *sb,
185 struct ext4_group_desc *bg)
186 {
187 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
188 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
189 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
190 }
191
ext4_itable_unused_count(struct super_block * sb,struct ext4_group_desc * bg)192 __u32 ext4_itable_unused_count(struct super_block *sb,
193 struct ext4_group_desc *bg)
194 {
195 return le16_to_cpu(bg->bg_itable_unused_lo) |
196 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
197 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
198 }
199
ext4_block_bitmap_set(struct super_block * sb,struct ext4_group_desc * bg,ext4_fsblk_t blk)200 void ext4_block_bitmap_set(struct super_block *sb,
201 struct ext4_group_desc *bg, ext4_fsblk_t blk)
202 {
203 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
204 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
205 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
206 }
207
ext4_inode_bitmap_set(struct super_block * sb,struct ext4_group_desc * bg,ext4_fsblk_t blk)208 void ext4_inode_bitmap_set(struct super_block *sb,
209 struct ext4_group_desc *bg, ext4_fsblk_t blk)
210 {
211 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
212 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
213 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
214 }
215
ext4_inode_table_set(struct super_block * sb,struct ext4_group_desc * bg,ext4_fsblk_t blk)216 void ext4_inode_table_set(struct super_block *sb,
217 struct ext4_group_desc *bg, ext4_fsblk_t blk)
218 {
219 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
220 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
221 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
222 }
223
ext4_free_group_clusters_set(struct super_block * sb,struct ext4_group_desc * bg,__u32 count)224 void ext4_free_group_clusters_set(struct super_block *sb,
225 struct ext4_group_desc *bg, __u32 count)
226 {
227 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
228 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
229 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
230 }
231
ext4_free_inodes_set(struct super_block * sb,struct ext4_group_desc * bg,__u32 count)232 void ext4_free_inodes_set(struct super_block *sb,
233 struct ext4_group_desc *bg, __u32 count)
234 {
235 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
236 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
237 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
238 }
239
ext4_used_dirs_set(struct super_block * sb,struct ext4_group_desc * bg,__u32 count)240 void ext4_used_dirs_set(struct super_block *sb,
241 struct ext4_group_desc *bg, __u32 count)
242 {
243 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
244 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
245 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
246 }
247
ext4_itable_unused_set(struct super_block * sb,struct ext4_group_desc * bg,__u32 count)248 void ext4_itable_unused_set(struct super_block *sb,
249 struct ext4_group_desc *bg, __u32 count)
250 {
251 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
252 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
253 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
254 }
255
256
257 /* Just increment the non-pointer handle value */
ext4_get_nojournal(void)258 static handle_t *ext4_get_nojournal(void)
259 {
260 handle_t *handle = current->journal_info;
261 unsigned long ref_cnt = (unsigned long)handle;
262
263 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
264
265 ref_cnt++;
266 handle = (handle_t *)ref_cnt;
267
268 current->journal_info = handle;
269 return handle;
270 }
271
272
273 /* Decrement the non-pointer handle value */
ext4_put_nojournal(handle_t * handle)274 static void ext4_put_nojournal(handle_t *handle)
275 {
276 unsigned long ref_cnt = (unsigned long)handle;
277
278 BUG_ON(ref_cnt == 0);
279
280 ref_cnt--;
281 handle = (handle_t *)ref_cnt;
282
283 current->journal_info = handle;
284 }
285
286 /*
287 * Wrappers for jbd2_journal_start/end.
288 *
289 * The only special thing we need to do here is to make sure that all
290 * journal_end calls result in the superblock being marked dirty, so
291 * that sync() will call the filesystem's write_super callback if
292 * appropriate.
293 *
294 * To avoid j_barrier hold in userspace when a user calls freeze(),
295 * ext4 prevents a new handle from being started by s_frozen, which
296 * is in an upper layer.
297 */
ext4_journal_start_sb(struct super_block * sb,int nblocks)298 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
299 {
300 journal_t *journal;
301 handle_t *handle;
302
303 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
304 if (sb->s_flags & MS_RDONLY)
305 return ERR_PTR(-EROFS);
306
307 journal = EXT4_SB(sb)->s_journal;
308 handle = ext4_journal_current_handle();
309
310 /*
311 * If a handle has been started, it should be allowed to
312 * finish, otherwise deadlock could happen between freeze
313 * and others(e.g. truncate) due to the restart of the
314 * journal handle if the filesystem is forzen and active
315 * handles are not stopped.
316 */
317 if (!handle)
318 vfs_check_frozen(sb, SB_FREEZE_TRANS);
319
320 if (!journal)
321 return ext4_get_nojournal();
322 /*
323 * Special case here: if the journal has aborted behind our
324 * backs (eg. EIO in the commit thread), then we still need to
325 * take the FS itself readonly cleanly.
326 */
327 if (is_journal_aborted(journal)) {
328 ext4_abort(sb, "Detected aborted journal");
329 return ERR_PTR(-EROFS);
330 }
331 return jbd2_journal_start(journal, nblocks);
332 }
333
334 /*
335 * The only special thing we need to do here is to make sure that all
336 * jbd2_journal_stop calls result in the superblock being marked dirty, so
337 * that sync() will call the filesystem's write_super callback if
338 * appropriate.
339 */
__ext4_journal_stop(const char * where,unsigned int line,handle_t * handle)340 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
341 {
342 struct super_block *sb;
343 int err;
344 int rc;
345
346 if (!ext4_handle_valid(handle)) {
347 ext4_put_nojournal(handle);
348 return 0;
349 }
350 sb = handle->h_transaction->t_journal->j_private;
351 err = handle->h_err;
352 rc = jbd2_journal_stop(handle);
353
354 if (!err)
355 err = rc;
356 if (err)
357 __ext4_std_error(sb, where, line, err);
358 return err;
359 }
360
ext4_journal_abort_handle(const char * caller,unsigned int line,const char * err_fn,struct buffer_head * bh,handle_t * handle,int err)361 void ext4_journal_abort_handle(const char *caller, unsigned int line,
362 const char *err_fn, struct buffer_head *bh,
363 handle_t *handle, int err)
364 {
365 char nbuf[16];
366 const char *errstr = ext4_decode_error(NULL, err, nbuf);
367
368 BUG_ON(!ext4_handle_valid(handle));
369
370 if (bh)
371 BUFFER_TRACE(bh, "abort");
372
373 if (!handle->h_err)
374 handle->h_err = err;
375
376 if (is_handle_aborted(handle))
377 return;
378
379 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
380 caller, line, errstr, err_fn);
381
382 jbd2_journal_abort_handle(handle);
383 }
384
__save_error_info(struct super_block * sb,const char * func,unsigned int line)385 static void __save_error_info(struct super_block *sb, const char *func,
386 unsigned int line)
387 {
388 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
389
390 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
391 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
392 es->s_last_error_time = cpu_to_le32(get_seconds());
393 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
394 es->s_last_error_line = cpu_to_le32(line);
395 if (!es->s_first_error_time) {
396 es->s_first_error_time = es->s_last_error_time;
397 strncpy(es->s_first_error_func, func,
398 sizeof(es->s_first_error_func));
399 es->s_first_error_line = cpu_to_le32(line);
400 es->s_first_error_ino = es->s_last_error_ino;
401 es->s_first_error_block = es->s_last_error_block;
402 }
403 /*
404 * Start the daily error reporting function if it hasn't been
405 * started already
406 */
407 if (!es->s_error_count)
408 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
409 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
410 }
411
save_error_info(struct super_block * sb,const char * func,unsigned int line)412 static void save_error_info(struct super_block *sb, const char *func,
413 unsigned int line)
414 {
415 __save_error_info(sb, func, line);
416 ext4_commit_super(sb, 1);
417 }
418
419 /*
420 * The del_gendisk() function uninitializes the disk-specific data
421 * structures, including the bdi structure, without telling anyone
422 * else. Once this happens, any attempt to call mark_buffer_dirty()
423 * (for example, by ext4_commit_super), will cause a kernel OOPS.
424 * This is a kludge to prevent these oops until we can put in a proper
425 * hook in del_gendisk() to inform the VFS and file system layers.
426 */
block_device_ejected(struct super_block * sb)427 static int block_device_ejected(struct super_block *sb)
428 {
429 struct inode *bd_inode = sb->s_bdev->bd_inode;
430 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
431
432 return bdi->dev == NULL;
433 }
434
ext4_journal_commit_callback(journal_t * journal,transaction_t * txn)435 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
436 {
437 struct super_block *sb = journal->j_private;
438 struct ext4_sb_info *sbi = EXT4_SB(sb);
439 int error = is_journal_aborted(journal);
440 struct ext4_journal_cb_entry *jce;
441
442 BUG_ON(txn->t_state == T_FINISHED);
443 spin_lock(&sbi->s_md_lock);
444 while (!list_empty(&txn->t_private_list)) {
445 jce = list_entry(txn->t_private_list.next,
446 struct ext4_journal_cb_entry, jce_list);
447 list_del_init(&jce->jce_list);
448 spin_unlock(&sbi->s_md_lock);
449 jce->jce_func(sb, jce, error);
450 spin_lock(&sbi->s_md_lock);
451 }
452 spin_unlock(&sbi->s_md_lock);
453 }
454
455 /* Deal with the reporting of failure conditions on a filesystem such as
456 * inconsistencies detected or read IO failures.
457 *
458 * On ext2, we can store the error state of the filesystem in the
459 * superblock. That is not possible on ext4, because we may have other
460 * write ordering constraints on the superblock which prevent us from
461 * writing it out straight away; and given that the journal is about to
462 * be aborted, we can't rely on the current, or future, transactions to
463 * write out the superblock safely.
464 *
465 * We'll just use the jbd2_journal_abort() error code to record an error in
466 * the journal instead. On recovery, the journal will complain about
467 * that error until we've noted it down and cleared it.
468 */
469
ext4_handle_error(struct super_block * sb)470 static void ext4_handle_error(struct super_block *sb)
471 {
472 if (sb->s_flags & MS_RDONLY)
473 return;
474
475 if (!test_opt(sb, ERRORS_CONT)) {
476 journal_t *journal = EXT4_SB(sb)->s_journal;
477
478 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
479 if (journal)
480 jbd2_journal_abort(journal, -EIO);
481 }
482 if (test_opt(sb, ERRORS_RO)) {
483 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
484 sb->s_flags |= MS_RDONLY;
485 }
486 if (test_opt(sb, ERRORS_PANIC))
487 panic("EXT4-fs (device %s): panic forced after error\n",
488 sb->s_id);
489 }
490
__ext4_error(struct super_block * sb,const char * function,unsigned int line,const char * fmt,...)491 void __ext4_error(struct super_block *sb, const char *function,
492 unsigned int line, const char *fmt, ...)
493 {
494 struct va_format vaf;
495 va_list args;
496
497 va_start(args, fmt);
498 vaf.fmt = fmt;
499 vaf.va = &args;
500 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
501 sb->s_id, function, line, current->comm, &vaf);
502 va_end(args);
503 save_error_info(sb, function, line);
504
505 ext4_handle_error(sb);
506 }
507
ext4_error_inode(struct inode * inode,const char * function,unsigned int line,ext4_fsblk_t block,const char * fmt,...)508 void ext4_error_inode(struct inode *inode, const char *function,
509 unsigned int line, ext4_fsblk_t block,
510 const char *fmt, ...)
511 {
512 va_list args;
513 struct va_format vaf;
514 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
515
516 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
517 es->s_last_error_block = cpu_to_le64(block);
518 save_error_info(inode->i_sb, function, line);
519 va_start(args, fmt);
520 vaf.fmt = fmt;
521 vaf.va = &args;
522 if (block)
523 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
524 "inode #%lu: block %llu: comm %s: %pV\n",
525 inode->i_sb->s_id, function, line, inode->i_ino,
526 block, current->comm, &vaf);
527 else
528 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
529 "inode #%lu: comm %s: %pV\n",
530 inode->i_sb->s_id, function, line, inode->i_ino,
531 current->comm, &vaf);
532 va_end(args);
533
534 ext4_handle_error(inode->i_sb);
535 }
536
ext4_error_file(struct file * file,const char * function,unsigned int line,ext4_fsblk_t block,const char * fmt,...)537 void ext4_error_file(struct file *file, const char *function,
538 unsigned int line, ext4_fsblk_t block,
539 const char *fmt, ...)
540 {
541 va_list args;
542 struct va_format vaf;
543 struct ext4_super_block *es;
544 struct inode *inode = file->f_dentry->d_inode;
545 char pathname[80], *path;
546
547 es = EXT4_SB(inode->i_sb)->s_es;
548 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
549 save_error_info(inode->i_sb, function, line);
550 path = d_path(&(file->f_path), pathname, sizeof(pathname));
551 if (IS_ERR(path))
552 path = "(unknown)";
553 va_start(args, fmt);
554 vaf.fmt = fmt;
555 vaf.va = &args;
556 if (block)
557 printk(KERN_CRIT
558 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
559 "block %llu: comm %s: path %s: %pV\n",
560 inode->i_sb->s_id, function, line, inode->i_ino,
561 block, current->comm, path, &vaf);
562 else
563 printk(KERN_CRIT
564 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
565 "comm %s: path %s: %pV\n",
566 inode->i_sb->s_id, function, line, inode->i_ino,
567 current->comm, path, &vaf);
568 va_end(args);
569
570 ext4_handle_error(inode->i_sb);
571 }
572
ext4_decode_error(struct super_block * sb,int errno,char nbuf[16])573 static const char *ext4_decode_error(struct super_block *sb, int errno,
574 char nbuf[16])
575 {
576 char *errstr = NULL;
577
578 switch (errno) {
579 case -EIO:
580 errstr = "IO failure";
581 break;
582 case -ENOMEM:
583 errstr = "Out of memory";
584 break;
585 case -EROFS:
586 if (!sb || (EXT4_SB(sb)->s_journal &&
587 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
588 errstr = "Journal has aborted";
589 else
590 errstr = "Readonly filesystem";
591 break;
592 default:
593 /* If the caller passed in an extra buffer for unknown
594 * errors, textualise them now. Else we just return
595 * NULL. */
596 if (nbuf) {
597 /* Check for truncated error codes... */
598 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
599 errstr = nbuf;
600 }
601 break;
602 }
603
604 return errstr;
605 }
606
607 /* __ext4_std_error decodes expected errors from journaling functions
608 * automatically and invokes the appropriate error response. */
609
__ext4_std_error(struct super_block * sb,const char * function,unsigned int line,int errno)610 void __ext4_std_error(struct super_block *sb, const char *function,
611 unsigned int line, int errno)
612 {
613 char nbuf[16];
614 const char *errstr;
615
616 /* Special case: if the error is EROFS, and we're not already
617 * inside a transaction, then there's really no point in logging
618 * an error. */
619 if (errno == -EROFS && journal_current_handle() == NULL &&
620 (sb->s_flags & MS_RDONLY))
621 return;
622
623 errstr = ext4_decode_error(sb, errno, nbuf);
624 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
625 sb->s_id, function, line, errstr);
626 save_error_info(sb, function, line);
627
628 ext4_handle_error(sb);
629 }
630
631 /*
632 * ext4_abort is a much stronger failure handler than ext4_error. The
633 * abort function may be used to deal with unrecoverable failures such
634 * as journal IO errors or ENOMEM at a critical moment in log management.
635 *
636 * We unconditionally force the filesystem into an ABORT|READONLY state,
637 * unless the error response on the fs has been set to panic in which
638 * case we take the easy way out and panic immediately.
639 */
640
__ext4_abort(struct super_block * sb,const char * function,unsigned int line,const char * fmt,...)641 void __ext4_abort(struct super_block *sb, const char *function,
642 unsigned int line, const char *fmt, ...)
643 {
644 va_list args;
645
646 save_error_info(sb, function, line);
647 va_start(args, fmt);
648 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
649 function, line);
650 vprintk(fmt, args);
651 printk("\n");
652 va_end(args);
653
654 if ((sb->s_flags & MS_RDONLY) == 0) {
655 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
656 sb->s_flags |= MS_RDONLY;
657 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
658 if (EXT4_SB(sb)->s_journal)
659 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
660 save_error_info(sb, function, line);
661 }
662 if (test_opt(sb, ERRORS_PANIC))
663 panic("EXT4-fs panic from previous error\n");
664 }
665
ext4_msg(struct super_block * sb,const char * prefix,const char * fmt,...)666 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
667 {
668 struct va_format vaf;
669 va_list args;
670
671 va_start(args, fmt);
672 vaf.fmt = fmt;
673 vaf.va = &args;
674 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
675 va_end(args);
676 }
677
__ext4_warning(struct super_block * sb,const char * function,unsigned int line,const char * fmt,...)678 void __ext4_warning(struct super_block *sb, const char *function,
679 unsigned int line, const char *fmt, ...)
680 {
681 struct va_format vaf;
682 va_list args;
683
684 va_start(args, fmt);
685 vaf.fmt = fmt;
686 vaf.va = &args;
687 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
688 sb->s_id, function, line, &vaf);
689 va_end(args);
690 }
691
__ext4_grp_locked_error(const char * function,unsigned int line,struct super_block * sb,ext4_group_t grp,unsigned long ino,ext4_fsblk_t block,const char * fmt,...)692 void __ext4_grp_locked_error(const char *function, unsigned int line,
693 struct super_block *sb, ext4_group_t grp,
694 unsigned long ino, ext4_fsblk_t block,
695 const char *fmt, ...)
696 __releases(bitlock)
697 __acquires(bitlock)
698 {
699 struct va_format vaf;
700 va_list args;
701 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
702
703 es->s_last_error_ino = cpu_to_le32(ino);
704 es->s_last_error_block = cpu_to_le64(block);
705 __save_error_info(sb, function, line);
706
707 va_start(args, fmt);
708
709 vaf.fmt = fmt;
710 vaf.va = &args;
711 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
712 sb->s_id, function, line, grp);
713 if (ino)
714 printk(KERN_CONT "inode %lu: ", ino);
715 if (block)
716 printk(KERN_CONT "block %llu:", (unsigned long long) block);
717 printk(KERN_CONT "%pV\n", &vaf);
718 va_end(args);
719
720 if (test_opt(sb, ERRORS_CONT)) {
721 ext4_commit_super(sb, 0);
722 return;
723 }
724
725 ext4_unlock_group(sb, grp);
726 ext4_handle_error(sb);
727 /*
728 * We only get here in the ERRORS_RO case; relocking the group
729 * may be dangerous, but nothing bad will happen since the
730 * filesystem will have already been marked read/only and the
731 * journal has been aborted. We return 1 as a hint to callers
732 * who might what to use the return value from
733 * ext4_grp_locked_error() to distinguish between the
734 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
735 * aggressively from the ext4 function in question, with a
736 * more appropriate error code.
737 */
738 ext4_lock_group(sb, grp);
739 return;
740 }
741
ext4_update_dynamic_rev(struct super_block * sb)742 void ext4_update_dynamic_rev(struct super_block *sb)
743 {
744 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
745
746 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
747 return;
748
749 ext4_warning(sb,
750 "updating to rev %d because of new feature flag, "
751 "running e2fsck is recommended",
752 EXT4_DYNAMIC_REV);
753
754 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
755 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
756 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
757 /* leave es->s_feature_*compat flags alone */
758 /* es->s_uuid will be set by e2fsck if empty */
759
760 /*
761 * The rest of the superblock fields should be zero, and if not it
762 * means they are likely already in use, so leave them alone. We
763 * can leave it up to e2fsck to clean up any inconsistencies there.
764 */
765 }
766
767 /*
768 * Open the external journal device
769 */
ext4_blkdev_get(dev_t dev,struct super_block * sb)770 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
771 {
772 struct block_device *bdev;
773 char b[BDEVNAME_SIZE];
774
775 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
776 if (IS_ERR(bdev))
777 goto fail;
778 return bdev;
779
780 fail:
781 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
782 __bdevname(dev, b), PTR_ERR(bdev));
783 return NULL;
784 }
785
786 /*
787 * Release the journal device
788 */
ext4_blkdev_put(struct block_device * bdev)789 static int ext4_blkdev_put(struct block_device *bdev)
790 {
791 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
792 }
793
ext4_blkdev_remove(struct ext4_sb_info * sbi)794 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
795 {
796 struct block_device *bdev;
797 int ret = -ENODEV;
798
799 bdev = sbi->journal_bdev;
800 if (bdev) {
801 ret = ext4_blkdev_put(bdev);
802 sbi->journal_bdev = NULL;
803 }
804 return ret;
805 }
806
orphan_list_entry(struct list_head * l)807 static inline struct inode *orphan_list_entry(struct list_head *l)
808 {
809 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
810 }
811
dump_orphan_list(struct super_block * sb,struct ext4_sb_info * sbi)812 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
813 {
814 struct list_head *l;
815
816 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
817 le32_to_cpu(sbi->s_es->s_last_orphan));
818
819 printk(KERN_ERR "sb_info orphan list:\n");
820 list_for_each(l, &sbi->s_orphan) {
821 struct inode *inode = orphan_list_entry(l);
822 printk(KERN_ERR " "
823 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
824 inode->i_sb->s_id, inode->i_ino, inode,
825 inode->i_mode, inode->i_nlink,
826 NEXT_ORPHAN(inode));
827 }
828 }
829
ext4_put_super(struct super_block * sb)830 static void ext4_put_super(struct super_block *sb)
831 {
832 struct ext4_sb_info *sbi = EXT4_SB(sb);
833 struct ext4_super_block *es = sbi->s_es;
834 int i, err;
835
836 ext4_unregister_li_request(sb);
837 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
838
839 flush_workqueue(sbi->dio_unwritten_wq);
840 destroy_workqueue(sbi->dio_unwritten_wq);
841
842 lock_super(sb);
843 if (sbi->s_journal) {
844 err = jbd2_journal_destroy(sbi->s_journal);
845 sbi->s_journal = NULL;
846 if (err < 0)
847 ext4_abort(sb, "Couldn't clean up the journal");
848 }
849
850 del_timer(&sbi->s_err_report);
851 ext4_release_system_zone(sb);
852 ext4_mb_release(sb);
853 ext4_ext_release(sb);
854 ext4_xattr_put_super(sb);
855
856 if (!(sb->s_flags & MS_RDONLY)) {
857 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
858 es->s_state = cpu_to_le16(sbi->s_mount_state);
859 }
860 if (sb->s_dirt || !(sb->s_flags & MS_RDONLY))
861 ext4_commit_super(sb, 1);
862
863 if (sbi->s_proc) {
864 remove_proc_entry("options", sbi->s_proc);
865 remove_proc_entry(sb->s_id, ext4_proc_root);
866 }
867 kobject_del(&sbi->s_kobj);
868
869 for (i = 0; i < sbi->s_gdb_count; i++)
870 brelse(sbi->s_group_desc[i]);
871 ext4_kvfree(sbi->s_group_desc);
872 ext4_kvfree(sbi->s_flex_groups);
873 percpu_counter_destroy(&sbi->s_freeclusters_counter);
874 percpu_counter_destroy(&sbi->s_freeinodes_counter);
875 percpu_counter_destroy(&sbi->s_dirs_counter);
876 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
877 brelse(sbi->s_sbh);
878 #ifdef CONFIG_QUOTA
879 for (i = 0; i < MAXQUOTAS; i++)
880 kfree(sbi->s_qf_names[i]);
881 #endif
882
883 /* Debugging code just in case the in-memory inode orphan list
884 * isn't empty. The on-disk one can be non-empty if we've
885 * detected an error and taken the fs readonly, but the
886 * in-memory list had better be clean by this point. */
887 if (!list_empty(&sbi->s_orphan))
888 dump_orphan_list(sb, sbi);
889 J_ASSERT(list_empty(&sbi->s_orphan));
890
891 invalidate_bdev(sb->s_bdev);
892 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
893 /*
894 * Invalidate the journal device's buffers. We don't want them
895 * floating about in memory - the physical journal device may
896 * hotswapped, and it breaks the `ro-after' testing code.
897 */
898 sync_blockdev(sbi->journal_bdev);
899 invalidate_bdev(sbi->journal_bdev);
900 ext4_blkdev_remove(sbi);
901 }
902 if (sbi->s_mmp_tsk)
903 kthread_stop(sbi->s_mmp_tsk);
904 sb->s_fs_info = NULL;
905 /*
906 * Now that we are completely done shutting down the
907 * superblock, we need to actually destroy the kobject.
908 */
909 unlock_super(sb);
910 kobject_put(&sbi->s_kobj);
911 wait_for_completion(&sbi->s_kobj_unregister);
912 kfree(sbi->s_blockgroup_lock);
913 kfree(sbi);
914 }
915
916 static struct kmem_cache *ext4_inode_cachep;
917
918 /*
919 * Called inside transaction, so use GFP_NOFS
920 */
ext4_alloc_inode(struct super_block * sb)921 static struct inode *ext4_alloc_inode(struct super_block *sb)
922 {
923 struct ext4_inode_info *ei;
924
925 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
926 if (!ei)
927 return NULL;
928
929 ei->vfs_inode.i_version = 1;
930 ei->vfs_inode.i_data.writeback_index = 0;
931 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
932 INIT_LIST_HEAD(&ei->i_prealloc_list);
933 spin_lock_init(&ei->i_prealloc_lock);
934 ei->i_reserved_data_blocks = 0;
935 ei->i_reserved_meta_blocks = 0;
936 ei->i_allocated_meta_blocks = 0;
937 ei->i_da_metadata_calc_len = 0;
938 ei->i_da_metadata_calc_last_lblock = 0;
939 spin_lock_init(&(ei->i_block_reservation_lock));
940 #ifdef CONFIG_QUOTA
941 ei->i_reserved_quota = 0;
942 #endif
943 ei->jinode = NULL;
944 INIT_LIST_HEAD(&ei->i_completed_io_list);
945 spin_lock_init(&ei->i_completed_io_lock);
946 ei->cur_aio_dio = NULL;
947 ei->i_sync_tid = 0;
948 ei->i_datasync_tid = 0;
949 atomic_set(&ei->i_ioend_count, 0);
950 atomic_set(&ei->i_aiodio_unwritten, 0);
951
952 return &ei->vfs_inode;
953 }
954
ext4_drop_inode(struct inode * inode)955 static int ext4_drop_inode(struct inode *inode)
956 {
957 int drop = generic_drop_inode(inode);
958
959 trace_ext4_drop_inode(inode, drop);
960 return drop;
961 }
962
ext4_i_callback(struct rcu_head * head)963 static void ext4_i_callback(struct rcu_head *head)
964 {
965 struct inode *inode = container_of(head, struct inode, i_rcu);
966 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
967 }
968
ext4_destroy_inode(struct inode * inode)969 static void ext4_destroy_inode(struct inode *inode)
970 {
971 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
972 ext4_msg(inode->i_sb, KERN_ERR,
973 "Inode %lu (%p): orphan list check failed!",
974 inode->i_ino, EXT4_I(inode));
975 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
976 EXT4_I(inode), sizeof(struct ext4_inode_info),
977 true);
978 dump_stack();
979 }
980 call_rcu(&inode->i_rcu, ext4_i_callback);
981 }
982
init_once(void * foo)983 static void init_once(void *foo)
984 {
985 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
986
987 INIT_LIST_HEAD(&ei->i_orphan);
988 #ifdef CONFIG_EXT4_FS_XATTR
989 init_rwsem(&ei->xattr_sem);
990 #endif
991 init_rwsem(&ei->i_data_sem);
992 inode_init_once(&ei->vfs_inode);
993 }
994
init_inodecache(void)995 static int init_inodecache(void)
996 {
997 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
998 sizeof(struct ext4_inode_info),
999 0, (SLAB_RECLAIM_ACCOUNT|
1000 SLAB_MEM_SPREAD),
1001 init_once);
1002 if (ext4_inode_cachep == NULL)
1003 return -ENOMEM;
1004 return 0;
1005 }
1006
destroy_inodecache(void)1007 static void destroy_inodecache(void)
1008 {
1009 kmem_cache_destroy(ext4_inode_cachep);
1010 }
1011
ext4_clear_inode(struct inode * inode)1012 void ext4_clear_inode(struct inode *inode)
1013 {
1014 invalidate_inode_buffers(inode);
1015 end_writeback(inode);
1016 dquot_drop(inode);
1017 ext4_discard_preallocations(inode);
1018 if (EXT4_I(inode)->jinode) {
1019 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1020 EXT4_I(inode)->jinode);
1021 jbd2_free_inode(EXT4_I(inode)->jinode);
1022 EXT4_I(inode)->jinode = NULL;
1023 }
1024 }
1025
ext4_nfs_get_inode(struct super_block * sb,u64 ino,u32 generation)1026 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1027 u64 ino, u32 generation)
1028 {
1029 struct inode *inode;
1030
1031 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1032 return ERR_PTR(-ESTALE);
1033 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1034 return ERR_PTR(-ESTALE);
1035
1036 /* iget isn't really right if the inode is currently unallocated!!
1037 *
1038 * ext4_read_inode will return a bad_inode if the inode had been
1039 * deleted, so we should be safe.
1040 *
1041 * Currently we don't know the generation for parent directory, so
1042 * a generation of 0 means "accept any"
1043 */
1044 inode = ext4_iget(sb, ino);
1045 if (IS_ERR(inode))
1046 return ERR_CAST(inode);
1047 if (generation && inode->i_generation != generation) {
1048 iput(inode);
1049 return ERR_PTR(-ESTALE);
1050 }
1051
1052 return inode;
1053 }
1054
ext4_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)1055 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1056 int fh_len, int fh_type)
1057 {
1058 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1059 ext4_nfs_get_inode);
1060 }
1061
ext4_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)1062 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1063 int fh_len, int fh_type)
1064 {
1065 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1066 ext4_nfs_get_inode);
1067 }
1068
1069 /*
1070 * Try to release metadata pages (indirect blocks, directories) which are
1071 * mapped via the block device. Since these pages could have journal heads
1072 * which would prevent try_to_free_buffers() from freeing them, we must use
1073 * jbd2 layer's try_to_free_buffers() function to release them.
1074 */
bdev_try_to_free_page(struct super_block * sb,struct page * page,gfp_t wait)1075 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1076 gfp_t wait)
1077 {
1078 journal_t *journal = EXT4_SB(sb)->s_journal;
1079
1080 WARN_ON(PageChecked(page));
1081 if (!page_has_buffers(page))
1082 return 0;
1083 if (journal)
1084 return jbd2_journal_try_to_free_buffers(journal, page,
1085 wait & ~__GFP_WAIT);
1086 return try_to_free_buffers(page);
1087 }
1088
1089 #ifdef CONFIG_QUOTA
1090 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1091 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1092
1093 static int ext4_write_dquot(struct dquot *dquot);
1094 static int ext4_acquire_dquot(struct dquot *dquot);
1095 static int ext4_release_dquot(struct dquot *dquot);
1096 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1097 static int ext4_write_info(struct super_block *sb, int type);
1098 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1099 struct path *path);
1100 static int ext4_quota_off(struct super_block *sb, int type);
1101 static int ext4_quota_on_mount(struct super_block *sb, int type);
1102 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1103 size_t len, loff_t off);
1104 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1105 const char *data, size_t len, loff_t off);
1106
1107 static const struct dquot_operations ext4_quota_operations = {
1108 .get_reserved_space = ext4_get_reserved_space,
1109 .write_dquot = ext4_write_dquot,
1110 .acquire_dquot = ext4_acquire_dquot,
1111 .release_dquot = ext4_release_dquot,
1112 .mark_dirty = ext4_mark_dquot_dirty,
1113 .write_info = ext4_write_info,
1114 .alloc_dquot = dquot_alloc,
1115 .destroy_dquot = dquot_destroy,
1116 };
1117
1118 static const struct quotactl_ops ext4_qctl_operations = {
1119 .quota_on = ext4_quota_on,
1120 .quota_off = ext4_quota_off,
1121 .quota_sync = dquot_quota_sync,
1122 .get_info = dquot_get_dqinfo,
1123 .set_info = dquot_set_dqinfo,
1124 .get_dqblk = dquot_get_dqblk,
1125 .set_dqblk = dquot_set_dqblk
1126 };
1127 #endif
1128
1129 static const struct super_operations ext4_sops = {
1130 .alloc_inode = ext4_alloc_inode,
1131 .destroy_inode = ext4_destroy_inode,
1132 .write_inode = ext4_write_inode,
1133 .dirty_inode = ext4_dirty_inode,
1134 .drop_inode = ext4_drop_inode,
1135 .evict_inode = ext4_evict_inode,
1136 .put_super = ext4_put_super,
1137 .sync_fs = ext4_sync_fs,
1138 .freeze_fs = ext4_freeze,
1139 .unfreeze_fs = ext4_unfreeze,
1140 .statfs = ext4_statfs,
1141 .remount_fs = ext4_remount,
1142 .show_options = ext4_show_options,
1143 #ifdef CONFIG_QUOTA
1144 .quota_read = ext4_quota_read,
1145 .quota_write = ext4_quota_write,
1146 #endif
1147 .bdev_try_to_free_page = bdev_try_to_free_page,
1148 };
1149
1150 static const struct super_operations ext4_nojournal_sops = {
1151 .alloc_inode = ext4_alloc_inode,
1152 .destroy_inode = ext4_destroy_inode,
1153 .write_inode = ext4_write_inode,
1154 .dirty_inode = ext4_dirty_inode,
1155 .drop_inode = ext4_drop_inode,
1156 .evict_inode = ext4_evict_inode,
1157 .write_super = ext4_write_super,
1158 .put_super = ext4_put_super,
1159 .statfs = ext4_statfs,
1160 .remount_fs = ext4_remount,
1161 .show_options = ext4_show_options,
1162 #ifdef CONFIG_QUOTA
1163 .quota_read = ext4_quota_read,
1164 .quota_write = ext4_quota_write,
1165 #endif
1166 .bdev_try_to_free_page = bdev_try_to_free_page,
1167 };
1168
1169 static const struct export_operations ext4_export_ops = {
1170 .fh_to_dentry = ext4_fh_to_dentry,
1171 .fh_to_parent = ext4_fh_to_parent,
1172 .get_parent = ext4_get_parent,
1173 };
1174
1175 enum {
1176 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1177 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1178 Opt_nouid32, Opt_debug, Opt_removed,
1179 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1180 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1181 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1182 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1183 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1184 Opt_data_err_abort, Opt_data_err_ignore,
1185 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1186 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1187 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1188 Opt_usrquota, Opt_grpquota, Opt_i_version,
1189 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1190 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1191 Opt_inode_readahead_blks, Opt_journal_ioprio,
1192 Opt_dioread_nolock, Opt_dioread_lock,
1193 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1194 };
1195
1196 static const match_table_t tokens = {
1197 {Opt_bsd_df, "bsddf"},
1198 {Opt_minix_df, "minixdf"},
1199 {Opt_grpid, "grpid"},
1200 {Opt_grpid, "bsdgroups"},
1201 {Opt_nogrpid, "nogrpid"},
1202 {Opt_nogrpid, "sysvgroups"},
1203 {Opt_resgid, "resgid=%u"},
1204 {Opt_resuid, "resuid=%u"},
1205 {Opt_sb, "sb=%u"},
1206 {Opt_err_cont, "errors=continue"},
1207 {Opt_err_panic, "errors=panic"},
1208 {Opt_err_ro, "errors=remount-ro"},
1209 {Opt_nouid32, "nouid32"},
1210 {Opt_debug, "debug"},
1211 {Opt_removed, "oldalloc"},
1212 {Opt_removed, "orlov"},
1213 {Opt_user_xattr, "user_xattr"},
1214 {Opt_nouser_xattr, "nouser_xattr"},
1215 {Opt_acl, "acl"},
1216 {Opt_noacl, "noacl"},
1217 {Opt_noload, "norecovery"},
1218 {Opt_noload, "noload"},
1219 {Opt_removed, "nobh"},
1220 {Opt_removed, "bh"},
1221 {Opt_commit, "commit=%u"},
1222 {Opt_min_batch_time, "min_batch_time=%u"},
1223 {Opt_max_batch_time, "max_batch_time=%u"},
1224 {Opt_journal_dev, "journal_dev=%u"},
1225 {Opt_journal_checksum, "journal_checksum"},
1226 {Opt_journal_async_commit, "journal_async_commit"},
1227 {Opt_abort, "abort"},
1228 {Opt_data_journal, "data=journal"},
1229 {Opt_data_ordered, "data=ordered"},
1230 {Opt_data_writeback, "data=writeback"},
1231 {Opt_data_err_abort, "data_err=abort"},
1232 {Opt_data_err_ignore, "data_err=ignore"},
1233 {Opt_offusrjquota, "usrjquota="},
1234 {Opt_usrjquota, "usrjquota=%s"},
1235 {Opt_offgrpjquota, "grpjquota="},
1236 {Opt_grpjquota, "grpjquota=%s"},
1237 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1238 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1239 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1240 {Opt_grpquota, "grpquota"},
1241 {Opt_noquota, "noquota"},
1242 {Opt_quota, "quota"},
1243 {Opt_usrquota, "usrquota"},
1244 {Opt_barrier, "barrier=%u"},
1245 {Opt_barrier, "barrier"},
1246 {Opt_nobarrier, "nobarrier"},
1247 {Opt_i_version, "i_version"},
1248 {Opt_stripe, "stripe=%u"},
1249 {Opt_delalloc, "delalloc"},
1250 {Opt_nodelalloc, "nodelalloc"},
1251 {Opt_mblk_io_submit, "mblk_io_submit"},
1252 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1253 {Opt_block_validity, "block_validity"},
1254 {Opt_noblock_validity, "noblock_validity"},
1255 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1256 {Opt_journal_ioprio, "journal_ioprio=%u"},
1257 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1258 {Opt_auto_da_alloc, "auto_da_alloc"},
1259 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1260 {Opt_dioread_nolock, "dioread_nolock"},
1261 {Opt_dioread_lock, "dioread_lock"},
1262 {Opt_discard, "discard"},
1263 {Opt_nodiscard, "nodiscard"},
1264 {Opt_init_itable, "init_itable=%u"},
1265 {Opt_init_itable, "init_itable"},
1266 {Opt_noinit_itable, "noinit_itable"},
1267 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1268 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1269 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1270 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1271 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1272 {Opt_err, NULL},
1273 };
1274
get_sb_block(void ** data)1275 static ext4_fsblk_t get_sb_block(void **data)
1276 {
1277 ext4_fsblk_t sb_block;
1278 char *options = (char *) *data;
1279
1280 if (!options || strncmp(options, "sb=", 3) != 0)
1281 return 1; /* Default location */
1282
1283 options += 3;
1284 /* TODO: use simple_strtoll with >32bit ext4 */
1285 sb_block = simple_strtoul(options, &options, 0);
1286 if (*options && *options != ',') {
1287 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1288 (char *) *data);
1289 return 1;
1290 }
1291 if (*options == ',')
1292 options++;
1293 *data = (void *) options;
1294
1295 return sb_block;
1296 }
1297
1298 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1299 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1300 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1301
1302 #ifdef CONFIG_QUOTA
set_qf_name(struct super_block * sb,int qtype,substring_t * args)1303 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1304 {
1305 struct ext4_sb_info *sbi = EXT4_SB(sb);
1306 char *qname;
1307
1308 if (sb_any_quota_loaded(sb) &&
1309 !sbi->s_qf_names[qtype]) {
1310 ext4_msg(sb, KERN_ERR,
1311 "Cannot change journaled "
1312 "quota options when quota turned on");
1313 return -1;
1314 }
1315 qname = match_strdup(args);
1316 if (!qname) {
1317 ext4_msg(sb, KERN_ERR,
1318 "Not enough memory for storing quotafile name");
1319 return -1;
1320 }
1321 if (sbi->s_qf_names[qtype] &&
1322 strcmp(sbi->s_qf_names[qtype], qname)) {
1323 ext4_msg(sb, KERN_ERR,
1324 "%s quota file already specified", QTYPE2NAME(qtype));
1325 kfree(qname);
1326 return -1;
1327 }
1328 sbi->s_qf_names[qtype] = qname;
1329 if (strchr(sbi->s_qf_names[qtype], '/')) {
1330 ext4_msg(sb, KERN_ERR,
1331 "quotafile must be on filesystem root");
1332 kfree(sbi->s_qf_names[qtype]);
1333 sbi->s_qf_names[qtype] = NULL;
1334 return -1;
1335 }
1336 set_opt(sb, QUOTA);
1337 return 1;
1338 }
1339
clear_qf_name(struct super_block * sb,int qtype)1340 static int clear_qf_name(struct super_block *sb, int qtype)
1341 {
1342
1343 struct ext4_sb_info *sbi = EXT4_SB(sb);
1344
1345 if (sb_any_quota_loaded(sb) &&
1346 sbi->s_qf_names[qtype]) {
1347 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1348 " when quota turned on");
1349 return -1;
1350 }
1351 /*
1352 * The space will be released later when all options are confirmed
1353 * to be correct
1354 */
1355 sbi->s_qf_names[qtype] = NULL;
1356 return 1;
1357 }
1358 #endif
1359
1360 #define MOPT_SET 0x0001
1361 #define MOPT_CLEAR 0x0002
1362 #define MOPT_NOSUPPORT 0x0004
1363 #define MOPT_EXPLICIT 0x0008
1364 #define MOPT_CLEAR_ERR 0x0010
1365 #define MOPT_GTE0 0x0020
1366 #ifdef CONFIG_QUOTA
1367 #define MOPT_Q 0
1368 #define MOPT_QFMT 0x0040
1369 #else
1370 #define MOPT_Q MOPT_NOSUPPORT
1371 #define MOPT_QFMT MOPT_NOSUPPORT
1372 #endif
1373 #define MOPT_DATAJ 0x0080
1374
1375 static const struct mount_opts {
1376 int token;
1377 int mount_opt;
1378 int flags;
1379 } ext4_mount_opts[] = {
1380 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1381 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1382 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1383 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1384 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1385 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1386 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1387 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1388 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1389 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1390 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1391 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1392 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1393 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1394 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1395 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1396 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1397 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1398 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1399 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1400 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1401 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1402 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1403 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1404 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1405 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1406 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1407 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1408 {Opt_commit, 0, MOPT_GTE0},
1409 {Opt_max_batch_time, 0, MOPT_GTE0},
1410 {Opt_min_batch_time, 0, MOPT_GTE0},
1411 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1412 {Opt_init_itable, 0, MOPT_GTE0},
1413 {Opt_stripe, 0, MOPT_GTE0},
1414 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1415 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1416 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
1417 #ifdef CONFIG_EXT4_FS_XATTR
1418 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1419 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1420 #else
1421 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1422 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
1423 #endif
1424 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1425 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1426 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1427 #else
1428 {Opt_acl, 0, MOPT_NOSUPPORT},
1429 {Opt_noacl, 0, MOPT_NOSUPPORT},
1430 #endif
1431 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1432 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1433 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1434 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1435 MOPT_SET | MOPT_Q},
1436 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1437 MOPT_SET | MOPT_Q},
1438 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1439 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1440 {Opt_usrjquota, 0, MOPT_Q},
1441 {Opt_grpjquota, 0, MOPT_Q},
1442 {Opt_offusrjquota, 0, MOPT_Q},
1443 {Opt_offgrpjquota, 0, MOPT_Q},
1444 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1445 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1446 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1447 {Opt_err, 0, 0}
1448 };
1449
handle_mount_opt(struct super_block * sb,char * opt,int token,substring_t * args,unsigned long * journal_devnum,unsigned int * journal_ioprio,int is_remount)1450 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1451 substring_t *args, unsigned long *journal_devnum,
1452 unsigned int *journal_ioprio, int is_remount)
1453 {
1454 struct ext4_sb_info *sbi = EXT4_SB(sb);
1455 const struct mount_opts *m;
1456 int arg = 0;
1457
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
1468 if (args->from && match_int(args, &arg))
1469 return -1;
1470 switch (token) {
1471 case Opt_noacl:
1472 case Opt_nouser_xattr:
1473 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1474 break;
1475 case Opt_sb:
1476 return 1; /* handled by get_sb_block() */
1477 case Opt_removed:
1478 ext4_msg(sb, KERN_WARNING,
1479 "Ignoring removed %s option", opt);
1480 return 1;
1481 case Opt_resuid:
1482 sbi->s_resuid = arg;
1483 return 1;
1484 case Opt_resgid:
1485 sbi->s_resgid = arg;
1486 return 1;
1487 case Opt_abort:
1488 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1489 return 1;
1490 case Opt_i_version:
1491 sb->s_flags |= MS_I_VERSION;
1492 return 1;
1493 case Opt_journal_dev:
1494 if (is_remount) {
1495 ext4_msg(sb, KERN_ERR,
1496 "Cannot specify journal on remount");
1497 return -1;
1498 }
1499 *journal_devnum = arg;
1500 return 1;
1501 case Opt_journal_ioprio:
1502 if (arg < 0 || arg > 7)
1503 return -1;
1504 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1505 return 1;
1506 }
1507
1508 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1509 if (token != m->token)
1510 continue;
1511 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1512 return -1;
1513 if (m->flags & MOPT_EXPLICIT)
1514 set_opt2(sb, EXPLICIT_DELALLOC);
1515 if (m->flags & MOPT_CLEAR_ERR)
1516 clear_opt(sb, ERRORS_MASK);
1517 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1518 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1519 "options when quota turned on");
1520 return -1;
1521 }
1522
1523 if (m->flags & MOPT_NOSUPPORT) {
1524 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1525 } else if (token == Opt_commit) {
1526 if (arg == 0)
1527 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1528 sbi->s_commit_interval = HZ * arg;
1529 } else if (token == Opt_max_batch_time) {
1530 if (arg == 0)
1531 arg = EXT4_DEF_MAX_BATCH_TIME;
1532 sbi->s_max_batch_time = arg;
1533 } else if (token == Opt_min_batch_time) {
1534 sbi->s_min_batch_time = arg;
1535 } else if (token == Opt_inode_readahead_blks) {
1536 if (arg > (1 << 30))
1537 return -1;
1538 if (arg && !is_power_of_2(arg)) {
1539 ext4_msg(sb, KERN_ERR,
1540 "EXT4-fs: inode_readahead_blks"
1541 " must be a power of 2");
1542 return -1;
1543 }
1544 sbi->s_inode_readahead_blks = arg;
1545 } else if (token == Opt_init_itable) {
1546 set_opt(sb, INIT_INODE_TABLE);
1547 if (!args->from)
1548 arg = EXT4_DEF_LI_WAIT_MULT;
1549 sbi->s_li_wait_mult = arg;
1550 } else if (token == Opt_stripe) {
1551 sbi->s_stripe = arg;
1552 } else if (m->flags & MOPT_DATAJ) {
1553 if (is_remount) {
1554 if (!sbi->s_journal)
1555 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1556 else if (test_opt(sb, DATA_FLAGS) !=
1557 m->mount_opt) {
1558 ext4_msg(sb, KERN_ERR,
1559 "Cannot change data mode on remount");
1560 return -1;
1561 }
1562 } else {
1563 clear_opt(sb, DATA_FLAGS);
1564 sbi->s_mount_opt |= m->mount_opt;
1565 }
1566 #ifdef CONFIG_QUOTA
1567 } else if (m->flags & MOPT_QFMT) {
1568 if (sb_any_quota_loaded(sb) &&
1569 sbi->s_jquota_fmt != m->mount_opt) {
1570 ext4_msg(sb, KERN_ERR, "Cannot "
1571 "change journaled quota options "
1572 "when quota turned on");
1573 return -1;
1574 }
1575 sbi->s_jquota_fmt = m->mount_opt;
1576 #endif
1577 } else {
1578 if (!args->from)
1579 arg = 1;
1580 if (m->flags & MOPT_CLEAR)
1581 arg = !arg;
1582 else if (unlikely(!(m->flags & MOPT_SET))) {
1583 ext4_msg(sb, KERN_WARNING,
1584 "buggy handling of option %s", opt);
1585 WARN_ON(1);
1586 return -1;
1587 }
1588 if (arg != 0)
1589 sbi->s_mount_opt |= m->mount_opt;
1590 else
1591 sbi->s_mount_opt &= ~m->mount_opt;
1592 }
1593 return 1;
1594 }
1595 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1596 "or missing value", opt);
1597 return -1;
1598 }
1599
parse_options(char * options,struct super_block * sb,unsigned long * journal_devnum,unsigned int * journal_ioprio,int is_remount)1600 static int parse_options(char *options, struct super_block *sb,
1601 unsigned long *journal_devnum,
1602 unsigned int *journal_ioprio,
1603 int is_remount)
1604 {
1605 struct ext4_sb_info *sbi = EXT4_SB(sb);
1606 char *p;
1607 substring_t args[MAX_OPT_ARGS];
1608 int token;
1609
1610 if (!options)
1611 return 1;
1612
1613 while ((p = strsep(&options, ",")) != NULL) {
1614 if (!*p)
1615 continue;
1616 /*
1617 * Initialize args struct so we know whether arg was
1618 * found; some options take optional arguments.
1619 */
1620 args[0].to = args[0].from = 0;
1621 token = match_token(p, tokens, args);
1622 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1623 journal_ioprio, is_remount) < 0)
1624 return 0;
1625 }
1626 #ifdef CONFIG_QUOTA
1627 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1628 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1629 clear_opt(sb, USRQUOTA);
1630
1631 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1632 clear_opt(sb, GRPQUOTA);
1633
1634 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1635 ext4_msg(sb, KERN_ERR, "old and new quota "
1636 "format mixing");
1637 return 0;
1638 }
1639
1640 if (!sbi->s_jquota_fmt) {
1641 ext4_msg(sb, KERN_ERR, "journaled quota format "
1642 "not specified");
1643 return 0;
1644 }
1645 } else {
1646 if (sbi->s_jquota_fmt) {
1647 ext4_msg(sb, KERN_ERR, "journaled quota format "
1648 "specified with no journaling "
1649 "enabled");
1650 return 0;
1651 }
1652 }
1653 #endif
1654 if (test_opt(sb, DIOREAD_NOLOCK)) {
1655 int blocksize =
1656 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1657
1658 if (blocksize < PAGE_CACHE_SIZE) {
1659 ext4_msg(sb, KERN_ERR, "can't mount with "
1660 "dioread_nolock if block size != PAGE_SIZE");
1661 return 0;
1662 }
1663 }
1664 return 1;
1665 }
1666
ext4_show_quota_options(struct seq_file * seq,struct super_block * sb)1667 static inline void ext4_show_quota_options(struct seq_file *seq,
1668 struct super_block *sb)
1669 {
1670 #if defined(CONFIG_QUOTA)
1671 struct ext4_sb_info *sbi = EXT4_SB(sb);
1672
1673 if (sbi->s_jquota_fmt) {
1674 char *fmtname = "";
1675
1676 switch (sbi->s_jquota_fmt) {
1677 case QFMT_VFS_OLD:
1678 fmtname = "vfsold";
1679 break;
1680 case QFMT_VFS_V0:
1681 fmtname = "vfsv0";
1682 break;
1683 case QFMT_VFS_V1:
1684 fmtname = "vfsv1";
1685 break;
1686 }
1687 seq_printf(seq, ",jqfmt=%s", fmtname);
1688 }
1689
1690 if (sbi->s_qf_names[USRQUOTA])
1691 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1692
1693 if (sbi->s_qf_names[GRPQUOTA])
1694 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1695
1696 if (test_opt(sb, USRQUOTA))
1697 seq_puts(seq, ",usrquota");
1698
1699 if (test_opt(sb, GRPQUOTA))
1700 seq_puts(seq, ",grpquota");
1701 #endif
1702 }
1703
token2str(int token)1704 static const char *token2str(int token)
1705 {
1706 const struct match_token *t;
1707
1708 for (t = tokens; t->token != Opt_err; t++)
1709 if (t->token == token && !strchr(t->pattern, '='))
1710 break;
1711 return t->pattern;
1712 }
1713
1714 /*
1715 * Show an option if
1716 * - it's set to a non-default value OR
1717 * - if the per-sb default is different from the global default
1718 */
_ext4_show_options(struct seq_file * seq,struct super_block * sb,int nodefs)1719 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1720 int nodefs)
1721 {
1722 struct ext4_sb_info *sbi = EXT4_SB(sb);
1723 struct ext4_super_block *es = sbi->s_es;
1724 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1725 const struct mount_opts *m;
1726 char sep = nodefs ? '\n' : ',';
1727
1728 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1729 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1730
1731 if (sbi->s_sb_block != 1)
1732 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1733
1734 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1735 int want_set = m->flags & MOPT_SET;
1736 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1737 (m->flags & MOPT_CLEAR_ERR))
1738 continue;
1739 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1740 continue; /* skip if same as the default */
1741 if ((want_set &&
1742 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1743 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1744 continue; /* select Opt_noFoo vs Opt_Foo */
1745 SEQ_OPTS_PRINT("%s", token2str(m->token));
1746 }
1747
1748 if (nodefs || sbi->s_resuid != EXT4_DEF_RESUID ||
1749 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1750 SEQ_OPTS_PRINT("resuid=%u", sbi->s_resuid);
1751 if (nodefs || sbi->s_resgid != EXT4_DEF_RESGID ||
1752 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1753 SEQ_OPTS_PRINT("resgid=%u", sbi->s_resgid);
1754 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1755 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1756 SEQ_OPTS_PUTS("errors=remount-ro");
1757 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1758 SEQ_OPTS_PUTS("errors=continue");
1759 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1760 SEQ_OPTS_PUTS("errors=panic");
1761 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1762 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1763 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1764 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1765 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1766 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1767 if (sb->s_flags & MS_I_VERSION)
1768 SEQ_OPTS_PUTS("i_version");
1769 if (nodefs || sbi->s_stripe)
1770 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1771 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1772 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1773 SEQ_OPTS_PUTS("data=journal");
1774 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1775 SEQ_OPTS_PUTS("data=ordered");
1776 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1777 SEQ_OPTS_PUTS("data=writeback");
1778 }
1779 if (nodefs ||
1780 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1781 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1782 sbi->s_inode_readahead_blks);
1783
1784 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1785 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1786 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1787
1788 ext4_show_quota_options(seq, sb);
1789 return 0;
1790 }
1791
ext4_show_options(struct seq_file * seq,struct dentry * root)1792 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1793 {
1794 return _ext4_show_options(seq, root->d_sb, 0);
1795 }
1796
options_seq_show(struct seq_file * seq,void * offset)1797 static int options_seq_show(struct seq_file *seq, void *offset)
1798 {
1799 struct super_block *sb = seq->private;
1800 int rc;
1801
1802 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1803 rc = _ext4_show_options(seq, sb, 1);
1804 seq_puts(seq, "\n");
1805 return rc;
1806 }
1807
options_open_fs(struct inode * inode,struct file * file)1808 static int options_open_fs(struct inode *inode, struct file *file)
1809 {
1810 return single_open(file, options_seq_show, PDE(inode)->data);
1811 }
1812
1813 static const struct file_operations ext4_seq_options_fops = {
1814 .owner = THIS_MODULE,
1815 .open = options_open_fs,
1816 .read = seq_read,
1817 .llseek = seq_lseek,
1818 .release = single_release,
1819 };
1820
ext4_setup_super(struct super_block * sb,struct ext4_super_block * es,int read_only)1821 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1822 int read_only)
1823 {
1824 struct ext4_sb_info *sbi = EXT4_SB(sb);
1825 int res = 0;
1826
1827 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1828 ext4_msg(sb, KERN_ERR, "revision level too high, "
1829 "forcing read-only mode");
1830 res = MS_RDONLY;
1831 }
1832 if (read_only)
1833 goto done;
1834 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1835 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1836 "running e2fsck is recommended");
1837 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1838 ext4_msg(sb, KERN_WARNING,
1839 "warning: mounting fs with errors, "
1840 "running e2fsck is recommended");
1841 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1842 le16_to_cpu(es->s_mnt_count) >=
1843 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1844 ext4_msg(sb, KERN_WARNING,
1845 "warning: maximal mount count reached, "
1846 "running e2fsck is recommended");
1847 else if (le32_to_cpu(es->s_checkinterval) &&
1848 (le32_to_cpu(es->s_lastcheck) +
1849 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1850 ext4_msg(sb, KERN_WARNING,
1851 "warning: checktime reached, "
1852 "running e2fsck is recommended");
1853 if (!sbi->s_journal)
1854 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1855 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1856 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1857 le16_add_cpu(&es->s_mnt_count, 1);
1858 es->s_mtime = cpu_to_le32(get_seconds());
1859 ext4_update_dynamic_rev(sb);
1860 if (sbi->s_journal)
1861 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1862
1863 ext4_commit_super(sb, 1);
1864 done:
1865 if (test_opt(sb, DEBUG))
1866 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1867 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1868 sb->s_blocksize,
1869 sbi->s_groups_count,
1870 EXT4_BLOCKS_PER_GROUP(sb),
1871 EXT4_INODES_PER_GROUP(sb),
1872 sbi->s_mount_opt, sbi->s_mount_opt2);
1873
1874 cleancache_init_fs(sb);
1875 return res;
1876 }
1877
ext4_fill_flex_info(struct super_block * sb)1878 static int ext4_fill_flex_info(struct super_block *sb)
1879 {
1880 struct ext4_sb_info *sbi = EXT4_SB(sb);
1881 struct ext4_group_desc *gdp = NULL;
1882 ext4_group_t flex_group_count;
1883 ext4_group_t flex_group;
1884 unsigned int groups_per_flex = 0;
1885 size_t size;
1886 int i;
1887
1888 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1889 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1890 sbi->s_log_groups_per_flex = 0;
1891 return 1;
1892 }
1893 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1894
1895 /* We allocate both existing and potentially added groups */
1896 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1897 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1898 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1899 size = flex_group_count * sizeof(struct flex_groups);
1900 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
1901 if (sbi->s_flex_groups == NULL) {
1902 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1903 flex_group_count);
1904 goto failed;
1905 }
1906
1907 for (i = 0; i < sbi->s_groups_count; i++) {
1908 gdp = ext4_get_group_desc(sb, i, NULL);
1909
1910 flex_group = ext4_flex_group(sbi, i);
1911 atomic_add(ext4_free_inodes_count(sb, gdp),
1912 &sbi->s_flex_groups[flex_group].free_inodes);
1913 atomic64_add(ext4_free_group_clusters(sb, gdp),
1914 &sbi->s_flex_groups[flex_group].free_clusters);
1915 atomic_add(ext4_used_dirs_count(sb, gdp),
1916 &sbi->s_flex_groups[flex_group].used_dirs);
1917 }
1918
1919 return 1;
1920 failed:
1921 return 0;
1922 }
1923
ext4_group_desc_csum(struct ext4_sb_info * sbi,__u32 block_group,struct ext4_group_desc * gdp)1924 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1925 struct ext4_group_desc *gdp)
1926 {
1927 __u16 crc = 0;
1928
1929 if (sbi->s_es->s_feature_ro_compat &
1930 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1931 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1932 __le32 le_group = cpu_to_le32(block_group);
1933
1934 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1935 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1936 crc = crc16(crc, (__u8 *)gdp, offset);
1937 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1938 /* for checksum of struct ext4_group_desc do the rest...*/
1939 if ((sbi->s_es->s_feature_incompat &
1940 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1941 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1942 crc = crc16(crc, (__u8 *)gdp + offset,
1943 le16_to_cpu(sbi->s_es->s_desc_size) -
1944 offset);
1945 }
1946
1947 return cpu_to_le16(crc);
1948 }
1949
ext4_group_desc_csum_verify(struct ext4_sb_info * sbi,__u32 block_group,struct ext4_group_desc * gdp)1950 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1951 struct ext4_group_desc *gdp)
1952 {
1953 if ((sbi->s_es->s_feature_ro_compat &
1954 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1955 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1956 return 0;
1957
1958 return 1;
1959 }
1960
1961 /* Called at mount-time, super-block is locked */
ext4_check_descriptors(struct super_block * sb,ext4_group_t * first_not_zeroed)1962 static int ext4_check_descriptors(struct super_block *sb,
1963 ext4_group_t *first_not_zeroed)
1964 {
1965 struct ext4_sb_info *sbi = EXT4_SB(sb);
1966 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1967 ext4_fsblk_t last_block;
1968 ext4_fsblk_t block_bitmap;
1969 ext4_fsblk_t inode_bitmap;
1970 ext4_fsblk_t inode_table;
1971 int flexbg_flag = 0;
1972 ext4_group_t i, grp = sbi->s_groups_count;
1973
1974 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1975 flexbg_flag = 1;
1976
1977 ext4_debug("Checking group descriptors");
1978
1979 for (i = 0; i < sbi->s_groups_count; i++) {
1980 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1981
1982 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1983 last_block = ext4_blocks_count(sbi->s_es) - 1;
1984 else
1985 last_block = first_block +
1986 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1987
1988 if ((grp == sbi->s_groups_count) &&
1989 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
1990 grp = i;
1991
1992 block_bitmap = ext4_block_bitmap(sb, gdp);
1993 if (block_bitmap < first_block || block_bitmap > last_block) {
1994 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1995 "Block bitmap for group %u not in group "
1996 "(block %llu)!", i, block_bitmap);
1997 return 0;
1998 }
1999 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2000 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2001 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2002 "Inode bitmap for group %u not in group "
2003 "(block %llu)!", i, inode_bitmap);
2004 return 0;
2005 }
2006 inode_table = ext4_inode_table(sb, gdp);
2007 if (inode_table < first_block ||
2008 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2009 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2010 "Inode table for group %u not in group "
2011 "(block %llu)!", i, inode_table);
2012 return 0;
2013 }
2014 ext4_lock_group(sb, i);
2015 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2016 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2017 "Checksum for group %u failed (%u!=%u)",
2018 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2019 gdp)), le16_to_cpu(gdp->bg_checksum));
2020 if (!(sb->s_flags & MS_RDONLY)) {
2021 ext4_unlock_group(sb, i);
2022 return 0;
2023 }
2024 }
2025 ext4_unlock_group(sb, i);
2026 if (!flexbg_flag)
2027 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2028 }
2029 if (NULL != first_not_zeroed)
2030 *first_not_zeroed = grp;
2031
2032 ext4_free_blocks_count_set(sbi->s_es,
2033 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2034 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2035 return 1;
2036 }
2037
2038 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2039 * the superblock) which were deleted from all directories, but held open by
2040 * a process at the time of a crash. We walk the list and try to delete these
2041 * inodes at recovery time (only with a read-write filesystem).
2042 *
2043 * In order to keep the orphan inode chain consistent during traversal (in
2044 * case of crash during recovery), we link each inode into the superblock
2045 * orphan list_head and handle it the same way as an inode deletion during
2046 * normal operation (which journals the operations for us).
2047 *
2048 * We only do an iget() and an iput() on each inode, which is very safe if we
2049 * accidentally point at an in-use or already deleted inode. The worst that
2050 * can happen in this case is that we get a "bit already cleared" message from
2051 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2052 * e2fsck was run on this filesystem, and it must have already done the orphan
2053 * inode cleanup for us, so we can safely abort without any further action.
2054 */
ext4_orphan_cleanup(struct super_block * sb,struct ext4_super_block * es)2055 static void ext4_orphan_cleanup(struct super_block *sb,
2056 struct ext4_super_block *es)
2057 {
2058 unsigned int s_flags = sb->s_flags;
2059 int nr_orphans = 0, nr_truncates = 0;
2060 #ifdef CONFIG_QUOTA
2061 int i;
2062 #endif
2063 if (!es->s_last_orphan) {
2064 jbd_debug(4, "no orphan inodes to clean up\n");
2065 return;
2066 }
2067
2068 if (bdev_read_only(sb->s_bdev)) {
2069 ext4_msg(sb, KERN_ERR, "write access "
2070 "unavailable, skipping orphan cleanup");
2071 return;
2072 }
2073
2074 /* Check if feature set would not allow a r/w mount */
2075 if (!ext4_feature_set_ok(sb, 0)) {
2076 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2077 "unknown ROCOMPAT features");
2078 return;
2079 }
2080
2081 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2082 if (es->s_last_orphan)
2083 jbd_debug(1, "Errors on filesystem, "
2084 "clearing orphan list.\n");
2085 es->s_last_orphan = 0;
2086 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2087 return;
2088 }
2089
2090 if (s_flags & MS_RDONLY) {
2091 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2092 sb->s_flags &= ~MS_RDONLY;
2093 }
2094 #ifdef CONFIG_QUOTA
2095 /* Needed for iput() to work correctly and not trash data */
2096 sb->s_flags |= MS_ACTIVE;
2097 /* Turn on quotas so that they are updated correctly */
2098 for (i = 0; i < MAXQUOTAS; i++) {
2099 if (EXT4_SB(sb)->s_qf_names[i]) {
2100 int ret = ext4_quota_on_mount(sb, i);
2101 if (ret < 0)
2102 ext4_msg(sb, KERN_ERR,
2103 "Cannot turn on journaled "
2104 "quota: error %d", ret);
2105 }
2106 }
2107 #endif
2108
2109 while (es->s_last_orphan) {
2110 struct inode *inode;
2111
2112 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2113 if (IS_ERR(inode)) {
2114 es->s_last_orphan = 0;
2115 break;
2116 }
2117
2118 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2119 dquot_initialize(inode);
2120 if (inode->i_nlink) {
2121 ext4_msg(sb, KERN_DEBUG,
2122 "%s: truncating inode %lu to %lld bytes",
2123 __func__, inode->i_ino, inode->i_size);
2124 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2125 inode->i_ino, inode->i_size);
2126 mutex_lock(&inode->i_mutex);
2127 ext4_truncate(inode);
2128 mutex_unlock(&inode->i_mutex);
2129 nr_truncates++;
2130 } else {
2131 ext4_msg(sb, KERN_DEBUG,
2132 "%s: deleting unreferenced inode %lu",
2133 __func__, inode->i_ino);
2134 jbd_debug(2, "deleting unreferenced inode %lu\n",
2135 inode->i_ino);
2136 nr_orphans++;
2137 }
2138 iput(inode); /* The delete magic happens here! */
2139 }
2140
2141 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2142
2143 if (nr_orphans)
2144 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2145 PLURAL(nr_orphans));
2146 if (nr_truncates)
2147 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2148 PLURAL(nr_truncates));
2149 #ifdef CONFIG_QUOTA
2150 /* Turn quotas off */
2151 for (i = 0; i < MAXQUOTAS; i++) {
2152 if (sb_dqopt(sb)->files[i])
2153 dquot_quota_off(sb, i);
2154 }
2155 #endif
2156 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2157 }
2158
2159 /*
2160 * Maximal extent format file size.
2161 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2162 * extent format containers, within a sector_t, and within i_blocks
2163 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2164 * so that won't be a limiting factor.
2165 *
2166 * However there is other limiting factor. We do store extents in the form
2167 * of starting block and length, hence the resulting length of the extent
2168 * covering maximum file size must fit into on-disk format containers as
2169 * well. Given that length is always by 1 unit bigger than max unit (because
2170 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2171 *
2172 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2173 */
ext4_max_size(int blkbits,int has_huge_files)2174 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2175 {
2176 loff_t res;
2177 loff_t upper_limit = MAX_LFS_FILESIZE;
2178
2179 /* small i_blocks in vfs inode? */
2180 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2181 /*
2182 * CONFIG_LBDAF is not enabled implies the inode
2183 * i_block represent total blocks in 512 bytes
2184 * 32 == size of vfs inode i_blocks * 8
2185 */
2186 upper_limit = (1LL << 32) - 1;
2187
2188 /* total blocks in file system block size */
2189 upper_limit >>= (blkbits - 9);
2190 upper_limit <<= blkbits;
2191 }
2192
2193 /*
2194 * 32-bit extent-start container, ee_block. We lower the maxbytes
2195 * by one fs block, so ee_len can cover the extent of maximum file
2196 * size
2197 */
2198 res = (1LL << 32) - 1;
2199 res <<= blkbits;
2200
2201 /* Sanity check against vm- & vfs- imposed limits */
2202 if (res > upper_limit)
2203 res = upper_limit;
2204
2205 return res;
2206 }
2207
2208 /*
2209 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2210 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2211 * We need to be 1 filesystem block less than the 2^48 sector limit.
2212 */
ext4_max_bitmap_size(int bits,int has_huge_files)2213 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2214 {
2215 loff_t res = EXT4_NDIR_BLOCKS;
2216 int meta_blocks;
2217 loff_t upper_limit;
2218 /* This is calculated to be the largest file size for a dense, block
2219 * mapped file such that the file's total number of 512-byte sectors,
2220 * including data and all indirect blocks, does not exceed (2^48 - 1).
2221 *
2222 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2223 * number of 512-byte sectors of the file.
2224 */
2225
2226 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2227 /*
2228 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2229 * the inode i_block field represents total file blocks in
2230 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2231 */
2232 upper_limit = (1LL << 32) - 1;
2233
2234 /* total blocks in file system block size */
2235 upper_limit >>= (bits - 9);
2236
2237 } else {
2238 /*
2239 * We use 48 bit ext4_inode i_blocks
2240 * With EXT4_HUGE_FILE_FL set the i_blocks
2241 * represent total number of blocks in
2242 * file system block size
2243 */
2244 upper_limit = (1LL << 48) - 1;
2245
2246 }
2247
2248 /* indirect blocks */
2249 meta_blocks = 1;
2250 /* double indirect blocks */
2251 meta_blocks += 1 + (1LL << (bits-2));
2252 /* tripple indirect blocks */
2253 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2254
2255 upper_limit -= meta_blocks;
2256 upper_limit <<= bits;
2257
2258 res += 1LL << (bits-2);
2259 res += 1LL << (2*(bits-2));
2260 res += 1LL << (3*(bits-2));
2261 res <<= bits;
2262 if (res > upper_limit)
2263 res = upper_limit;
2264
2265 if (res > MAX_LFS_FILESIZE)
2266 res = MAX_LFS_FILESIZE;
2267
2268 return res;
2269 }
2270
descriptor_loc(struct super_block * sb,ext4_fsblk_t logical_sb_block,int nr)2271 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2272 ext4_fsblk_t logical_sb_block, int nr)
2273 {
2274 struct ext4_sb_info *sbi = EXT4_SB(sb);
2275 ext4_group_t bg, first_meta_bg;
2276 int has_super = 0;
2277
2278 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2279
2280 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2281 nr < first_meta_bg)
2282 return logical_sb_block + nr + 1;
2283 bg = sbi->s_desc_per_block * nr;
2284 if (ext4_bg_has_super(sb, bg))
2285 has_super = 1;
2286
2287 return (has_super + ext4_group_first_block_no(sb, bg));
2288 }
2289
2290 /**
2291 * ext4_get_stripe_size: Get the stripe size.
2292 * @sbi: In memory super block info
2293 *
2294 * If we have specified it via mount option, then
2295 * use the mount option value. If the value specified at mount time is
2296 * greater than the blocks per group use the super block value.
2297 * If the super block value is greater than blocks per group return 0.
2298 * Allocator needs it be less than blocks per group.
2299 *
2300 */
ext4_get_stripe_size(struct ext4_sb_info * sbi)2301 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2302 {
2303 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2304 unsigned long stripe_width =
2305 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2306 int ret;
2307
2308 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2309 ret = sbi->s_stripe;
2310 else if (stripe_width <= sbi->s_blocks_per_group)
2311 ret = stripe_width;
2312 else if (stride <= sbi->s_blocks_per_group)
2313 ret = stride;
2314 else
2315 ret = 0;
2316
2317 /*
2318 * If the stripe width is 1, this makes no sense and
2319 * we set it to 0 to turn off stripe handling code.
2320 */
2321 if (ret <= 1)
2322 ret = 0;
2323
2324 return ret;
2325 }
2326
2327 /* sysfs supprt */
2328
2329 struct ext4_attr {
2330 struct attribute attr;
2331 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2332 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2333 const char *, size_t);
2334 int offset;
2335 };
2336
parse_strtoul(const char * buf,unsigned long max,unsigned long * value)2337 static int parse_strtoul(const char *buf,
2338 unsigned long max, unsigned long *value)
2339 {
2340 char *endp;
2341
2342 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2343 endp = skip_spaces(endp);
2344 if (*endp || *value > max)
2345 return -EINVAL;
2346
2347 return 0;
2348 }
2349
delayed_allocation_blocks_show(struct ext4_attr * a,struct ext4_sb_info * sbi,char * buf)2350 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2351 struct ext4_sb_info *sbi,
2352 char *buf)
2353 {
2354 return snprintf(buf, PAGE_SIZE, "%llu\n",
2355 (s64) EXT4_C2B(sbi,
2356 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2357 }
2358
session_write_kbytes_show(struct ext4_attr * a,struct ext4_sb_info * sbi,char * buf)2359 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2360 struct ext4_sb_info *sbi, char *buf)
2361 {
2362 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2363
2364 if (!sb->s_bdev->bd_part)
2365 return snprintf(buf, PAGE_SIZE, "0\n");
2366 return snprintf(buf, PAGE_SIZE, "%lu\n",
2367 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2368 sbi->s_sectors_written_start) >> 1);
2369 }
2370
lifetime_write_kbytes_show(struct ext4_attr * a,struct ext4_sb_info * sbi,char * buf)2371 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2372 struct ext4_sb_info *sbi, char *buf)
2373 {
2374 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2375
2376 if (!sb->s_bdev->bd_part)
2377 return snprintf(buf, PAGE_SIZE, "0\n");
2378 return snprintf(buf, PAGE_SIZE, "%llu\n",
2379 (unsigned long long)(sbi->s_kbytes_written +
2380 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2381 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2382 }
2383
inode_readahead_blks_store(struct ext4_attr * a,struct ext4_sb_info * sbi,const char * buf,size_t count)2384 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2385 struct ext4_sb_info *sbi,
2386 const char *buf, size_t count)
2387 {
2388 unsigned long t;
2389
2390 if (parse_strtoul(buf, 0x40000000, &t))
2391 return -EINVAL;
2392
2393 if (t && !is_power_of_2(t))
2394 return -EINVAL;
2395
2396 sbi->s_inode_readahead_blks = t;
2397 return count;
2398 }
2399
sbi_ui_show(struct ext4_attr * a,struct ext4_sb_info * sbi,char * buf)2400 static ssize_t sbi_ui_show(struct ext4_attr *a,
2401 struct ext4_sb_info *sbi, char *buf)
2402 {
2403 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2404
2405 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2406 }
2407
sbi_ui_store(struct ext4_attr * a,struct ext4_sb_info * sbi,const char * buf,size_t count)2408 static ssize_t sbi_ui_store(struct ext4_attr *a,
2409 struct ext4_sb_info *sbi,
2410 const char *buf, size_t count)
2411 {
2412 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2413 unsigned long t;
2414
2415 if (parse_strtoul(buf, 0xffffffff, &t))
2416 return -EINVAL;
2417 *ui = t;
2418 return count;
2419 }
2420
2421 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2422 static struct ext4_attr ext4_attr_##_name = { \
2423 .attr = {.name = __stringify(_name), .mode = _mode }, \
2424 .show = _show, \
2425 .store = _store, \
2426 .offset = offsetof(struct ext4_sb_info, _elname), \
2427 }
2428 #define EXT4_ATTR(name, mode, show, store) \
2429 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2430
2431 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2432 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2433 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2434 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2435 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2436 #define ATTR_LIST(name) &ext4_attr_##name.attr
2437
2438 EXT4_RO_ATTR(delayed_allocation_blocks);
2439 EXT4_RO_ATTR(session_write_kbytes);
2440 EXT4_RO_ATTR(lifetime_write_kbytes);
2441 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2442 inode_readahead_blks_store, s_inode_readahead_blks);
2443 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2444 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2445 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2446 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2447 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2448 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2449 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2450 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2451
2452 static struct attribute *ext4_attrs[] = {
2453 ATTR_LIST(delayed_allocation_blocks),
2454 ATTR_LIST(session_write_kbytes),
2455 ATTR_LIST(lifetime_write_kbytes),
2456 ATTR_LIST(inode_readahead_blks),
2457 ATTR_LIST(inode_goal),
2458 ATTR_LIST(mb_stats),
2459 ATTR_LIST(mb_max_to_scan),
2460 ATTR_LIST(mb_min_to_scan),
2461 ATTR_LIST(mb_order2_req),
2462 ATTR_LIST(mb_stream_req),
2463 ATTR_LIST(mb_group_prealloc),
2464 ATTR_LIST(max_writeback_mb_bump),
2465 NULL,
2466 };
2467
2468 /* Features this copy of ext4 supports */
2469 EXT4_INFO_ATTR(lazy_itable_init);
2470 EXT4_INFO_ATTR(batched_discard);
2471
2472 static struct attribute *ext4_feat_attrs[] = {
2473 ATTR_LIST(lazy_itable_init),
2474 ATTR_LIST(batched_discard),
2475 NULL,
2476 };
2477
ext4_attr_show(struct kobject * kobj,struct attribute * attr,char * buf)2478 static ssize_t ext4_attr_show(struct kobject *kobj,
2479 struct attribute *attr, char *buf)
2480 {
2481 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2482 s_kobj);
2483 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2484
2485 return a->show ? a->show(a, sbi, buf) : 0;
2486 }
2487
ext4_attr_store(struct kobject * kobj,struct attribute * attr,const char * buf,size_t len)2488 static ssize_t ext4_attr_store(struct kobject *kobj,
2489 struct attribute *attr,
2490 const char *buf, size_t len)
2491 {
2492 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2493 s_kobj);
2494 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2495
2496 return a->store ? a->store(a, sbi, buf, len) : 0;
2497 }
2498
ext4_sb_release(struct kobject * kobj)2499 static void ext4_sb_release(struct kobject *kobj)
2500 {
2501 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2502 s_kobj);
2503 complete(&sbi->s_kobj_unregister);
2504 }
2505
2506 static const struct sysfs_ops ext4_attr_ops = {
2507 .show = ext4_attr_show,
2508 .store = ext4_attr_store,
2509 };
2510
2511 static struct kobj_type ext4_ktype = {
2512 .default_attrs = ext4_attrs,
2513 .sysfs_ops = &ext4_attr_ops,
2514 .release = ext4_sb_release,
2515 };
2516
ext4_feat_release(struct kobject * kobj)2517 static void ext4_feat_release(struct kobject *kobj)
2518 {
2519 complete(&ext4_feat->f_kobj_unregister);
2520 }
2521
2522 static struct kobj_type ext4_feat_ktype = {
2523 .default_attrs = ext4_feat_attrs,
2524 .sysfs_ops = &ext4_attr_ops,
2525 .release = ext4_feat_release,
2526 };
2527
2528 /*
2529 * Check whether this filesystem can be mounted based on
2530 * the features present and the RDONLY/RDWR mount requested.
2531 * Returns 1 if this filesystem can be mounted as requested,
2532 * 0 if it cannot be.
2533 */
ext4_feature_set_ok(struct super_block * sb,int readonly)2534 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2535 {
2536 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2537 ext4_msg(sb, KERN_ERR,
2538 "Couldn't mount because of "
2539 "unsupported optional features (%x)",
2540 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2541 ~EXT4_FEATURE_INCOMPAT_SUPP));
2542 return 0;
2543 }
2544
2545 if (readonly)
2546 return 1;
2547
2548 /* Check that feature set is OK for a read-write mount */
2549 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2550 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2551 "unsupported optional features (%x)",
2552 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2553 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2554 return 0;
2555 }
2556 /*
2557 * Large file size enabled file system can only be mounted
2558 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2559 */
2560 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2561 if (sizeof(blkcnt_t) < sizeof(u64)) {
2562 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2563 "cannot be mounted RDWR without "
2564 "CONFIG_LBDAF");
2565 return 0;
2566 }
2567 }
2568 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2569 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2570 ext4_msg(sb, KERN_ERR,
2571 "Can't support bigalloc feature without "
2572 "extents feature\n");
2573 return 0;
2574 }
2575 return 1;
2576 }
2577
2578 /*
2579 * This function is called once a day if we have errors logged
2580 * on the file system
2581 */
print_daily_error_info(unsigned long arg)2582 static void print_daily_error_info(unsigned long arg)
2583 {
2584 struct super_block *sb = (struct super_block *) arg;
2585 struct ext4_sb_info *sbi;
2586 struct ext4_super_block *es;
2587
2588 sbi = EXT4_SB(sb);
2589 es = sbi->s_es;
2590
2591 if (es->s_error_count)
2592 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2593 le32_to_cpu(es->s_error_count));
2594 if (es->s_first_error_time) {
2595 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2596 sb->s_id, le32_to_cpu(es->s_first_error_time),
2597 (int) sizeof(es->s_first_error_func),
2598 es->s_first_error_func,
2599 le32_to_cpu(es->s_first_error_line));
2600 if (es->s_first_error_ino)
2601 printk(": inode %u",
2602 le32_to_cpu(es->s_first_error_ino));
2603 if (es->s_first_error_block)
2604 printk(": block %llu", (unsigned long long)
2605 le64_to_cpu(es->s_first_error_block));
2606 printk("\n");
2607 }
2608 if (es->s_last_error_time) {
2609 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2610 sb->s_id, le32_to_cpu(es->s_last_error_time),
2611 (int) sizeof(es->s_last_error_func),
2612 es->s_last_error_func,
2613 le32_to_cpu(es->s_last_error_line));
2614 if (es->s_last_error_ino)
2615 printk(": inode %u",
2616 le32_to_cpu(es->s_last_error_ino));
2617 if (es->s_last_error_block)
2618 printk(": block %llu", (unsigned long long)
2619 le64_to_cpu(es->s_last_error_block));
2620 printk("\n");
2621 }
2622 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2623 }
2624
2625 /* Find next suitable group and run ext4_init_inode_table */
ext4_run_li_request(struct ext4_li_request * elr)2626 static int ext4_run_li_request(struct ext4_li_request *elr)
2627 {
2628 struct ext4_group_desc *gdp = NULL;
2629 ext4_group_t group, ngroups;
2630 struct super_block *sb;
2631 unsigned long timeout = 0;
2632 int ret = 0;
2633
2634 sb = elr->lr_super;
2635 ngroups = EXT4_SB(sb)->s_groups_count;
2636
2637 for (group = elr->lr_next_group; group < ngroups; group++) {
2638 gdp = ext4_get_group_desc(sb, group, NULL);
2639 if (!gdp) {
2640 ret = 1;
2641 break;
2642 }
2643
2644 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2645 break;
2646 }
2647
2648 if (group == ngroups)
2649 ret = 1;
2650
2651 if (!ret) {
2652 timeout = jiffies;
2653 ret = ext4_init_inode_table(sb, group,
2654 elr->lr_timeout ? 0 : 1);
2655 if (elr->lr_timeout == 0) {
2656 timeout = (jiffies - timeout) *
2657 elr->lr_sbi->s_li_wait_mult;
2658 elr->lr_timeout = timeout;
2659 }
2660 elr->lr_next_sched = jiffies + elr->lr_timeout;
2661 elr->lr_next_group = group + 1;
2662 }
2663
2664 return ret;
2665 }
2666
2667 /*
2668 * Remove lr_request from the list_request and free the
2669 * request structure. Should be called with li_list_mtx held
2670 */
ext4_remove_li_request(struct ext4_li_request * elr)2671 static void ext4_remove_li_request(struct ext4_li_request *elr)
2672 {
2673 struct ext4_sb_info *sbi;
2674
2675 if (!elr)
2676 return;
2677
2678 sbi = elr->lr_sbi;
2679
2680 list_del(&elr->lr_request);
2681 sbi->s_li_request = NULL;
2682 kfree(elr);
2683 }
2684
ext4_unregister_li_request(struct super_block * sb)2685 static void ext4_unregister_li_request(struct super_block *sb)
2686 {
2687 mutex_lock(&ext4_li_mtx);
2688 if (!ext4_li_info) {
2689 mutex_unlock(&ext4_li_mtx);
2690 return;
2691 }
2692
2693 mutex_lock(&ext4_li_info->li_list_mtx);
2694 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2695 mutex_unlock(&ext4_li_info->li_list_mtx);
2696 mutex_unlock(&ext4_li_mtx);
2697 }
2698
2699 static struct task_struct *ext4_lazyinit_task;
2700
2701 /*
2702 * This is the function where ext4lazyinit thread lives. It walks
2703 * through the request list searching for next scheduled filesystem.
2704 * When such a fs is found, run the lazy initialization request
2705 * (ext4_rn_li_request) and keep track of the time spend in this
2706 * function. Based on that time we compute next schedule time of
2707 * the request. When walking through the list is complete, compute
2708 * next waking time and put itself into sleep.
2709 */
ext4_lazyinit_thread(void * arg)2710 static int ext4_lazyinit_thread(void *arg)
2711 {
2712 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2713 struct list_head *pos, *n;
2714 struct ext4_li_request *elr;
2715 unsigned long next_wakeup, cur;
2716
2717 BUG_ON(NULL == eli);
2718
2719 cont_thread:
2720 while (true) {
2721 next_wakeup = MAX_JIFFY_OFFSET;
2722
2723 mutex_lock(&eli->li_list_mtx);
2724 if (list_empty(&eli->li_request_list)) {
2725 mutex_unlock(&eli->li_list_mtx);
2726 goto exit_thread;
2727 }
2728
2729 list_for_each_safe(pos, n, &eli->li_request_list) {
2730 elr = list_entry(pos, struct ext4_li_request,
2731 lr_request);
2732
2733 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2734 if (ext4_run_li_request(elr) != 0) {
2735 /* error, remove the lazy_init job */
2736 ext4_remove_li_request(elr);
2737 continue;
2738 }
2739 }
2740
2741 if (time_before(elr->lr_next_sched, next_wakeup))
2742 next_wakeup = elr->lr_next_sched;
2743 }
2744 mutex_unlock(&eli->li_list_mtx);
2745
2746 try_to_freeze();
2747
2748 cur = jiffies;
2749 if ((time_after_eq(cur, next_wakeup)) ||
2750 (MAX_JIFFY_OFFSET == next_wakeup)) {
2751 cond_resched();
2752 continue;
2753 }
2754
2755 schedule_timeout_interruptible(next_wakeup - cur);
2756
2757 if (kthread_should_stop()) {
2758 ext4_clear_request_list();
2759 goto exit_thread;
2760 }
2761 }
2762
2763 exit_thread:
2764 /*
2765 * It looks like the request list is empty, but we need
2766 * to check it under the li_list_mtx lock, to prevent any
2767 * additions into it, and of course we should lock ext4_li_mtx
2768 * to atomically free the list and ext4_li_info, because at
2769 * this point another ext4 filesystem could be registering
2770 * new one.
2771 */
2772 mutex_lock(&ext4_li_mtx);
2773 mutex_lock(&eli->li_list_mtx);
2774 if (!list_empty(&eli->li_request_list)) {
2775 mutex_unlock(&eli->li_list_mtx);
2776 mutex_unlock(&ext4_li_mtx);
2777 goto cont_thread;
2778 }
2779 mutex_unlock(&eli->li_list_mtx);
2780 kfree(ext4_li_info);
2781 ext4_li_info = NULL;
2782 mutex_unlock(&ext4_li_mtx);
2783
2784 return 0;
2785 }
2786
ext4_clear_request_list(void)2787 static void ext4_clear_request_list(void)
2788 {
2789 struct list_head *pos, *n;
2790 struct ext4_li_request *elr;
2791
2792 mutex_lock(&ext4_li_info->li_list_mtx);
2793 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2794 elr = list_entry(pos, struct ext4_li_request,
2795 lr_request);
2796 ext4_remove_li_request(elr);
2797 }
2798 mutex_unlock(&ext4_li_info->li_list_mtx);
2799 }
2800
ext4_run_lazyinit_thread(void)2801 static int ext4_run_lazyinit_thread(void)
2802 {
2803 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2804 ext4_li_info, "ext4lazyinit");
2805 if (IS_ERR(ext4_lazyinit_task)) {
2806 int err = PTR_ERR(ext4_lazyinit_task);
2807 ext4_clear_request_list();
2808 kfree(ext4_li_info);
2809 ext4_li_info = NULL;
2810 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2811 "initialization thread\n",
2812 err);
2813 return err;
2814 }
2815 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2816 return 0;
2817 }
2818
2819 /*
2820 * Check whether it make sense to run itable init. thread or not.
2821 * If there is at least one uninitialized inode table, return
2822 * corresponding group number, else the loop goes through all
2823 * groups and return total number of groups.
2824 */
ext4_has_uninit_itable(struct super_block * sb)2825 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2826 {
2827 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2828 struct ext4_group_desc *gdp = NULL;
2829
2830 for (group = 0; group < ngroups; group++) {
2831 gdp = ext4_get_group_desc(sb, group, NULL);
2832 if (!gdp)
2833 continue;
2834
2835 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2836 break;
2837 }
2838
2839 return group;
2840 }
2841
ext4_li_info_new(void)2842 static int ext4_li_info_new(void)
2843 {
2844 struct ext4_lazy_init *eli = NULL;
2845
2846 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2847 if (!eli)
2848 return -ENOMEM;
2849
2850 INIT_LIST_HEAD(&eli->li_request_list);
2851 mutex_init(&eli->li_list_mtx);
2852
2853 eli->li_state |= EXT4_LAZYINIT_QUIT;
2854
2855 ext4_li_info = eli;
2856
2857 return 0;
2858 }
2859
ext4_li_request_new(struct super_block * sb,ext4_group_t start)2860 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2861 ext4_group_t start)
2862 {
2863 struct ext4_sb_info *sbi = EXT4_SB(sb);
2864 struct ext4_li_request *elr;
2865 unsigned long rnd;
2866
2867 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2868 if (!elr)
2869 return NULL;
2870
2871 elr->lr_super = sb;
2872 elr->lr_sbi = sbi;
2873 elr->lr_next_group = start;
2874
2875 /*
2876 * Randomize first schedule time of the request to
2877 * spread the inode table initialization requests
2878 * better.
2879 */
2880 get_random_bytes(&rnd, sizeof(rnd));
2881 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2882 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2883
2884 return elr;
2885 }
2886
ext4_register_li_request(struct super_block * sb,ext4_group_t first_not_zeroed)2887 static int ext4_register_li_request(struct super_block *sb,
2888 ext4_group_t first_not_zeroed)
2889 {
2890 struct ext4_sb_info *sbi = EXT4_SB(sb);
2891 struct ext4_li_request *elr;
2892 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2893 int ret = 0;
2894
2895 if (sbi->s_li_request != NULL) {
2896 /*
2897 * Reset timeout so it can be computed again, because
2898 * s_li_wait_mult might have changed.
2899 */
2900 sbi->s_li_request->lr_timeout = 0;
2901 return 0;
2902 }
2903
2904 if (first_not_zeroed == ngroups ||
2905 (sb->s_flags & MS_RDONLY) ||
2906 !test_opt(sb, INIT_INODE_TABLE))
2907 return 0;
2908
2909 elr = ext4_li_request_new(sb, first_not_zeroed);
2910 if (!elr)
2911 return -ENOMEM;
2912
2913 mutex_lock(&ext4_li_mtx);
2914
2915 if (NULL == ext4_li_info) {
2916 ret = ext4_li_info_new();
2917 if (ret)
2918 goto out;
2919 }
2920
2921 mutex_lock(&ext4_li_info->li_list_mtx);
2922 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2923 mutex_unlock(&ext4_li_info->li_list_mtx);
2924
2925 sbi->s_li_request = elr;
2926 /*
2927 * set elr to NULL here since it has been inserted to
2928 * the request_list and the removal and free of it is
2929 * handled by ext4_clear_request_list from now on.
2930 */
2931 elr = NULL;
2932
2933 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2934 ret = ext4_run_lazyinit_thread();
2935 if (ret)
2936 goto out;
2937 }
2938 out:
2939 mutex_unlock(&ext4_li_mtx);
2940 if (ret)
2941 kfree(elr);
2942 return ret;
2943 }
2944
2945 /*
2946 * We do not need to lock anything since this is called on
2947 * module unload.
2948 */
ext4_destroy_lazyinit_thread(void)2949 static void ext4_destroy_lazyinit_thread(void)
2950 {
2951 /*
2952 * If thread exited earlier
2953 * there's nothing to be done.
2954 */
2955 if (!ext4_li_info || !ext4_lazyinit_task)
2956 return;
2957
2958 kthread_stop(ext4_lazyinit_task);
2959 }
2960
2961 /*
2962 * Note: calculating the overhead so we can be compatible with
2963 * historical BSD practice is quite difficult in the face of
2964 * clusters/bigalloc. This is because multiple metadata blocks from
2965 * different block group can end up in the same allocation cluster.
2966 * Calculating the exact overhead in the face of clustered allocation
2967 * requires either O(all block bitmaps) in memory or O(number of block
2968 * groups**2) in time. We will still calculate the superblock for
2969 * older file systems --- and if we come across with a bigalloc file
2970 * system with zero in s_overhead_clusters the estimate will be close to
2971 * correct especially for very large cluster sizes --- but for newer
2972 * file systems, it's better to calculate this figure once at mkfs
2973 * time, and store it in the superblock. If the superblock value is
2974 * present (even for non-bigalloc file systems), we will use it.
2975 */
count_overhead(struct super_block * sb,ext4_group_t grp,char * buf)2976 static int count_overhead(struct super_block *sb, ext4_group_t grp,
2977 char *buf)
2978 {
2979 struct ext4_sb_info *sbi = EXT4_SB(sb);
2980 struct ext4_group_desc *gdp;
2981 ext4_fsblk_t first_block, last_block, b;
2982 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
2983 int s, j, count = 0;
2984
2985 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
2986 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
2987 sbi->s_itb_per_group + 2);
2988
2989 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
2990 (grp * EXT4_BLOCKS_PER_GROUP(sb));
2991 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
2992 for (i = 0; i < ngroups; i++) {
2993 gdp = ext4_get_group_desc(sb, i, NULL);
2994 b = ext4_block_bitmap(sb, gdp);
2995 if (b >= first_block && b <= last_block) {
2996 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
2997 count++;
2998 }
2999 b = ext4_inode_bitmap(sb, gdp);
3000 if (b >= first_block && b <= last_block) {
3001 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3002 count++;
3003 }
3004 b = ext4_inode_table(sb, gdp);
3005 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3006 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3007 int c = EXT4_B2C(sbi, b - first_block);
3008 ext4_set_bit(c, buf);
3009 count++;
3010 }
3011 if (i != grp)
3012 continue;
3013 s = 0;
3014 if (ext4_bg_has_super(sb, grp)) {
3015 ext4_set_bit(s++, buf);
3016 count++;
3017 }
3018 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3019 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3020 count++;
3021 }
3022 }
3023 if (!count)
3024 return 0;
3025 return EXT4_CLUSTERS_PER_GROUP(sb) -
3026 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3027 }
3028
3029 /*
3030 * Compute the overhead and stash it in sbi->s_overhead
3031 */
ext4_calculate_overhead(struct super_block * sb)3032 int ext4_calculate_overhead(struct super_block *sb)
3033 {
3034 struct ext4_sb_info *sbi = EXT4_SB(sb);
3035 struct ext4_super_block *es = sbi->s_es;
3036 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3037 ext4_fsblk_t overhead = 0;
3038 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3039
3040 memset(buf, 0, PAGE_SIZE);
3041 if (!buf)
3042 return -ENOMEM;
3043
3044 /*
3045 * Compute the overhead (FS structures). This is constant
3046 * for a given filesystem unless the number of block groups
3047 * changes so we cache the previous value until it does.
3048 */
3049
3050 /*
3051 * All of the blocks before first_data_block are overhead
3052 */
3053 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3054
3055 /*
3056 * Add the overhead found in each block group
3057 */
3058 for (i = 0; i < ngroups; i++) {
3059 int blks;
3060
3061 blks = count_overhead(sb, i, buf);
3062 overhead += blks;
3063 if (blks)
3064 memset(buf, 0, PAGE_SIZE);
3065 cond_resched();
3066 }
3067 sbi->s_overhead = overhead;
3068 smp_wmb();
3069 free_page((unsigned long) buf);
3070 return 0;
3071 }
3072
ext4_fill_super(struct super_block * sb,void * data,int silent)3073 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3074 {
3075 char *orig_data = kstrdup(data, GFP_KERNEL);
3076 struct buffer_head *bh;
3077 struct ext4_super_block *es = NULL;
3078 struct ext4_sb_info *sbi;
3079 ext4_fsblk_t block;
3080 ext4_fsblk_t sb_block = get_sb_block(&data);
3081 ext4_fsblk_t logical_sb_block;
3082 unsigned long offset = 0;
3083 unsigned long journal_devnum = 0;
3084 unsigned long def_mount_opts;
3085 struct inode *root;
3086 char *cp;
3087 const char *descr;
3088 int ret = -ENOMEM;
3089 int blocksize, clustersize;
3090 unsigned int db_count;
3091 unsigned int i;
3092 int needs_recovery, has_huge_files, has_bigalloc;
3093 __u64 blocks_count;
3094 int err;
3095 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3096 ext4_group_t first_not_zeroed;
3097
3098 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3099 if (!sbi)
3100 goto out_free_orig;
3101
3102 sbi->s_blockgroup_lock =
3103 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3104 if (!sbi->s_blockgroup_lock) {
3105 kfree(sbi);
3106 goto out_free_orig;
3107 }
3108 sb->s_fs_info = sbi;
3109 sbi->s_mount_opt = 0;
3110 sbi->s_resuid = EXT4_DEF_RESUID;
3111 sbi->s_resgid = EXT4_DEF_RESGID;
3112 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3113 sbi->s_sb_block = sb_block;
3114 if (sb->s_bdev->bd_part)
3115 sbi->s_sectors_written_start =
3116 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3117
3118 /* Cleanup superblock name */
3119 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3120 *cp = '!';
3121
3122 ret = -EINVAL;
3123 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3124 if (!blocksize) {
3125 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3126 goto out_fail;
3127 }
3128
3129 /*
3130 * The ext4 superblock will not be buffer aligned for other than 1kB
3131 * block sizes. We need to calculate the offset from buffer start.
3132 */
3133 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3134 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3135 offset = do_div(logical_sb_block, blocksize);
3136 } else {
3137 logical_sb_block = sb_block;
3138 }
3139
3140 if (!(bh = sb_bread(sb, logical_sb_block))) {
3141 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3142 goto out_fail;
3143 }
3144 /*
3145 * Note: s_es must be initialized as soon as possible because
3146 * some ext4 macro-instructions depend on its value
3147 */
3148 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3149 sbi->s_es = es;
3150 sb->s_magic = le16_to_cpu(es->s_magic);
3151 if (sb->s_magic != EXT4_SUPER_MAGIC)
3152 goto cantfind_ext4;
3153 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3154
3155 /* Set defaults before we parse the mount options */
3156 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3157 set_opt(sb, INIT_INODE_TABLE);
3158 if (def_mount_opts & EXT4_DEFM_DEBUG)
3159 set_opt(sb, DEBUG);
3160 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3161 set_opt(sb, GRPID);
3162 if (def_mount_opts & EXT4_DEFM_UID16)
3163 set_opt(sb, NO_UID32);
3164 /* xattr user namespace & acls are now defaulted on */
3165 #ifdef CONFIG_EXT4_FS_XATTR
3166 set_opt(sb, XATTR_USER);
3167 #endif
3168 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3169 set_opt(sb, POSIX_ACL);
3170 #endif
3171 set_opt(sb, MBLK_IO_SUBMIT);
3172 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3173 set_opt(sb, JOURNAL_DATA);
3174 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3175 set_opt(sb, ORDERED_DATA);
3176 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3177 set_opt(sb, WRITEBACK_DATA);
3178
3179 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3180 set_opt(sb, ERRORS_PANIC);
3181 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3182 set_opt(sb, ERRORS_CONT);
3183 else
3184 set_opt(sb, ERRORS_RO);
3185 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3186 set_opt(sb, BLOCK_VALIDITY);
3187 if (def_mount_opts & EXT4_DEFM_DISCARD)
3188 set_opt(sb, DISCARD);
3189
3190 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3191 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3192 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3193 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3194 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3195
3196 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3197 set_opt(sb, BARRIER);
3198
3199 /*
3200 * enable delayed allocation by default
3201 * Use -o nodelalloc to turn it off
3202 */
3203 if (!IS_EXT3_SB(sb) &&
3204 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3205 set_opt(sb, DELALLOC);
3206
3207 /*
3208 * set default s_li_wait_mult for lazyinit, for the case there is
3209 * no mount option specified.
3210 */
3211 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3212
3213 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3214 &journal_devnum, &journal_ioprio, 0)) {
3215 ext4_msg(sb, KERN_WARNING,
3216 "failed to parse options in superblock: %s",
3217 sbi->s_es->s_mount_opts);
3218 }
3219 sbi->s_def_mount_opt = sbi->s_mount_opt;
3220 if (!parse_options((char *) data, sb, &journal_devnum,
3221 &journal_ioprio, 0))
3222 goto failed_mount;
3223
3224 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3225 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3226 "with data=journal disables delayed "
3227 "allocation and O_DIRECT support!\n");
3228 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3229 ext4_msg(sb, KERN_ERR, "can't mount with "
3230 "both data=journal and delalloc");
3231 goto failed_mount;
3232 }
3233 if (test_opt(sb, DIOREAD_NOLOCK)) {
3234 ext4_msg(sb, KERN_ERR, "can't mount with "
3235 "both data=journal and dioread_nolock");
3236 goto failed_mount;
3237 }
3238 if (test_opt(sb, DELALLOC))
3239 clear_opt(sb, DELALLOC);
3240 }
3241
3242 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3243 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3244
3245 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3246 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3247 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3248 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3249 ext4_msg(sb, KERN_WARNING,
3250 "feature flags set on rev 0 fs, "
3251 "running e2fsck is recommended");
3252
3253 if (IS_EXT2_SB(sb)) {
3254 if (ext2_feature_set_ok(sb))
3255 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3256 "using the ext4 subsystem");
3257 else {
3258 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3259 "to feature incompatibilities");
3260 goto failed_mount;
3261 }
3262 }
3263
3264 if (IS_EXT3_SB(sb)) {
3265 if (ext3_feature_set_ok(sb))
3266 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3267 "using the ext4 subsystem");
3268 else {
3269 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3270 "to feature incompatibilities");
3271 goto failed_mount;
3272 }
3273 }
3274
3275 /*
3276 * Check feature flags regardless of the revision level, since we
3277 * previously didn't change the revision level when setting the flags,
3278 * so there is a chance incompat flags are set on a rev 0 filesystem.
3279 */
3280 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3281 goto failed_mount;
3282
3283 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3284 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3285 blocksize > EXT4_MAX_BLOCK_SIZE) {
3286 ext4_msg(sb, KERN_ERR,
3287 "Unsupported filesystem blocksize %d", blocksize);
3288 goto failed_mount;
3289 }
3290
3291 if (sb->s_blocksize != blocksize) {
3292 /* Validate the filesystem blocksize */
3293 if (!sb_set_blocksize(sb, blocksize)) {
3294 ext4_msg(sb, KERN_ERR, "bad block size %d",
3295 blocksize);
3296 goto failed_mount;
3297 }
3298
3299 brelse(bh);
3300 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3301 offset = do_div(logical_sb_block, blocksize);
3302 bh = sb_bread(sb, logical_sb_block);
3303 if (!bh) {
3304 ext4_msg(sb, KERN_ERR,
3305 "Can't read superblock on 2nd try");
3306 goto failed_mount;
3307 }
3308 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3309 sbi->s_es = es;
3310 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3311 ext4_msg(sb, KERN_ERR,
3312 "Magic mismatch, very weird!");
3313 goto failed_mount;
3314 }
3315 }
3316
3317 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3318 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3319 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3320 has_huge_files);
3321 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3322
3323 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3324 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3325 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3326 } else {
3327 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3328 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3329 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3330 (!is_power_of_2(sbi->s_inode_size)) ||
3331 (sbi->s_inode_size > blocksize)) {
3332 ext4_msg(sb, KERN_ERR,
3333 "unsupported inode size: %d",
3334 sbi->s_inode_size);
3335 goto failed_mount;
3336 }
3337 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3338 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3339 }
3340
3341 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3342 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3343 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3344 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3345 !is_power_of_2(sbi->s_desc_size)) {
3346 ext4_msg(sb, KERN_ERR,
3347 "unsupported descriptor size %lu",
3348 sbi->s_desc_size);
3349 goto failed_mount;
3350 }
3351 } else
3352 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3353
3354 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3355 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3356 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3357 goto cantfind_ext4;
3358
3359 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3360 if (sbi->s_inodes_per_block == 0)
3361 goto cantfind_ext4;
3362 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3363 sbi->s_inodes_per_block;
3364 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3365 sbi->s_sbh = bh;
3366 sbi->s_mount_state = le16_to_cpu(es->s_state);
3367 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3368 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3369
3370 for (i = 0; i < 4; i++)
3371 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3372 sbi->s_def_hash_version = es->s_def_hash_version;
3373 i = le32_to_cpu(es->s_flags);
3374 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3375 sbi->s_hash_unsigned = 3;
3376 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3377 #ifdef __CHAR_UNSIGNED__
3378 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3379 sbi->s_hash_unsigned = 3;
3380 #else
3381 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3382 #endif
3383 }
3384
3385 /* Handle clustersize */
3386 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3387 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3388 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3389 if (has_bigalloc) {
3390 if (clustersize < blocksize) {
3391 ext4_msg(sb, KERN_ERR,
3392 "cluster size (%d) smaller than "
3393 "block size (%d)", clustersize, blocksize);
3394 goto failed_mount;
3395 }
3396 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3397 le32_to_cpu(es->s_log_block_size);
3398 sbi->s_clusters_per_group =
3399 le32_to_cpu(es->s_clusters_per_group);
3400 if (sbi->s_clusters_per_group > blocksize * 8) {
3401 ext4_msg(sb, KERN_ERR,
3402 "#clusters per group too big: %lu",
3403 sbi->s_clusters_per_group);
3404 goto failed_mount;
3405 }
3406 if (sbi->s_blocks_per_group !=
3407 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3408 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3409 "clusters per group (%lu) inconsistent",
3410 sbi->s_blocks_per_group,
3411 sbi->s_clusters_per_group);
3412 goto failed_mount;
3413 }
3414 } else {
3415 if (clustersize != blocksize) {
3416 ext4_warning(sb, "fragment/cluster size (%d) != "
3417 "block size (%d)", clustersize,
3418 blocksize);
3419 clustersize = blocksize;
3420 }
3421 if (sbi->s_blocks_per_group > blocksize * 8) {
3422 ext4_msg(sb, KERN_ERR,
3423 "#blocks per group too big: %lu",
3424 sbi->s_blocks_per_group);
3425 goto failed_mount;
3426 }
3427 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3428 sbi->s_cluster_bits = 0;
3429 }
3430 sbi->s_cluster_ratio = clustersize / blocksize;
3431
3432 if (sbi->s_inodes_per_group > blocksize * 8) {
3433 ext4_msg(sb, KERN_ERR,
3434 "#inodes per group too big: %lu",
3435 sbi->s_inodes_per_group);
3436 goto failed_mount;
3437 }
3438
3439 /*
3440 * Test whether we have more sectors than will fit in sector_t,
3441 * and whether the max offset is addressable by the page cache.
3442 */
3443 err = generic_check_addressable(sb->s_blocksize_bits,
3444 ext4_blocks_count(es));
3445 if (err) {
3446 ext4_msg(sb, KERN_ERR, "filesystem"
3447 " too large to mount safely on this system");
3448 if (sizeof(sector_t) < 8)
3449 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3450 ret = err;
3451 goto failed_mount;
3452 }
3453
3454 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3455 goto cantfind_ext4;
3456
3457 /* check blocks count against device size */
3458 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3459 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3460 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3461 "exceeds size of device (%llu blocks)",
3462 ext4_blocks_count(es), blocks_count);
3463 goto failed_mount;
3464 }
3465
3466 /*
3467 * It makes no sense for the first data block to be beyond the end
3468 * of the filesystem.
3469 */
3470 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3471 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3472 "block %u is beyond end of filesystem (%llu)",
3473 le32_to_cpu(es->s_first_data_block),
3474 ext4_blocks_count(es));
3475 goto failed_mount;
3476 }
3477 blocks_count = (ext4_blocks_count(es) -
3478 le32_to_cpu(es->s_first_data_block) +
3479 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3480 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3481 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3482 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3483 "(block count %llu, first data block %u, "
3484 "blocks per group %lu)", sbi->s_groups_count,
3485 ext4_blocks_count(es),
3486 le32_to_cpu(es->s_first_data_block),
3487 EXT4_BLOCKS_PER_GROUP(sb));
3488 goto failed_mount;
3489 }
3490 sbi->s_groups_count = blocks_count;
3491 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3492 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3493 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3494 EXT4_DESC_PER_BLOCK(sb);
3495 sbi->s_group_desc = ext4_kvmalloc(db_count *
3496 sizeof(struct buffer_head *),
3497 GFP_KERNEL);
3498 if (sbi->s_group_desc == NULL) {
3499 ext4_msg(sb, KERN_ERR, "not enough memory");
3500 goto failed_mount;
3501 }
3502
3503 if (ext4_proc_root)
3504 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3505
3506 if (sbi->s_proc)
3507 proc_create_data("options", S_IRUGO, sbi->s_proc,
3508 &ext4_seq_options_fops, sb);
3509
3510 bgl_lock_init(sbi->s_blockgroup_lock);
3511
3512 for (i = 0; i < db_count; i++) {
3513 block = descriptor_loc(sb, logical_sb_block, i);
3514 sbi->s_group_desc[i] = sb_bread(sb, block);
3515 if (!sbi->s_group_desc[i]) {
3516 ext4_msg(sb, KERN_ERR,
3517 "can't read group descriptor %d", i);
3518 db_count = i;
3519 goto failed_mount2;
3520 }
3521 }
3522 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3523 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3524 goto failed_mount2;
3525 }
3526 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3527 if (!ext4_fill_flex_info(sb)) {
3528 ext4_msg(sb, KERN_ERR,
3529 "unable to initialize "
3530 "flex_bg meta info!");
3531 goto failed_mount2;
3532 }
3533
3534 sbi->s_gdb_count = db_count;
3535 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3536 spin_lock_init(&sbi->s_next_gen_lock);
3537
3538 init_timer(&sbi->s_err_report);
3539 sbi->s_err_report.function = print_daily_error_info;
3540 sbi->s_err_report.data = (unsigned long) sb;
3541
3542 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3543 ext4_count_free_clusters(sb));
3544 if (!err) {
3545 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3546 ext4_count_free_inodes(sb));
3547 }
3548 if (!err) {
3549 err = percpu_counter_init(&sbi->s_dirs_counter,
3550 ext4_count_dirs(sb));
3551 }
3552 if (!err) {
3553 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3554 }
3555 if (err) {
3556 ext4_msg(sb, KERN_ERR, "insufficient memory");
3557 goto failed_mount3;
3558 }
3559
3560 sbi->s_stripe = ext4_get_stripe_size(sbi);
3561 sbi->s_max_writeback_mb_bump = 128;
3562
3563 /*
3564 * set up enough so that it can read an inode
3565 */
3566 if (!test_opt(sb, NOLOAD) &&
3567 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3568 sb->s_op = &ext4_sops;
3569 else
3570 sb->s_op = &ext4_nojournal_sops;
3571 sb->s_export_op = &ext4_export_ops;
3572 sb->s_xattr = ext4_xattr_handlers;
3573 #ifdef CONFIG_QUOTA
3574 sb->s_qcop = &ext4_qctl_operations;
3575 sb->dq_op = &ext4_quota_operations;
3576 #endif
3577 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3578
3579 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3580 mutex_init(&sbi->s_orphan_lock);
3581 sbi->s_resize_flags = 0;
3582
3583 sb->s_root = NULL;
3584
3585 needs_recovery = (es->s_last_orphan != 0 ||
3586 EXT4_HAS_INCOMPAT_FEATURE(sb,
3587 EXT4_FEATURE_INCOMPAT_RECOVER));
3588
3589 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3590 !(sb->s_flags & MS_RDONLY))
3591 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3592 goto failed_mount3;
3593
3594 /*
3595 * The first inode we look at is the journal inode. Don't try
3596 * root first: it may be modified in the journal!
3597 */
3598 if (!test_opt(sb, NOLOAD) &&
3599 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3600 if (ext4_load_journal(sb, es, journal_devnum))
3601 goto failed_mount3;
3602 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3603 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3604 ext4_msg(sb, KERN_ERR, "required journal recovery "
3605 "suppressed and not mounted read-only");
3606 goto failed_mount_wq;
3607 } else {
3608 clear_opt(sb, DATA_FLAGS);
3609 sbi->s_journal = NULL;
3610 needs_recovery = 0;
3611 goto no_journal;
3612 }
3613
3614 if (ext4_blocks_count(es) > 0xffffffffULL &&
3615 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3616 JBD2_FEATURE_INCOMPAT_64BIT)) {
3617 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3618 goto failed_mount_wq;
3619 }
3620
3621 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3622 jbd2_journal_set_features(sbi->s_journal,
3623 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3624 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3625 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3626 jbd2_journal_set_features(sbi->s_journal,
3627 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3628 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3629 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3630 } else {
3631 jbd2_journal_clear_features(sbi->s_journal,
3632 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3633 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3634 }
3635
3636 /* We have now updated the journal if required, so we can
3637 * validate the data journaling mode. */
3638 switch (test_opt(sb, DATA_FLAGS)) {
3639 case 0:
3640 /* No mode set, assume a default based on the journal
3641 * capabilities: ORDERED_DATA if the journal can
3642 * cope, else JOURNAL_DATA
3643 */
3644 if (jbd2_journal_check_available_features
3645 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3646 set_opt(sb, ORDERED_DATA);
3647 else
3648 set_opt(sb, JOURNAL_DATA);
3649 break;
3650
3651 case EXT4_MOUNT_ORDERED_DATA:
3652 case EXT4_MOUNT_WRITEBACK_DATA:
3653 if (!jbd2_journal_check_available_features
3654 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3655 ext4_msg(sb, KERN_ERR, "Journal does not support "
3656 "requested data journaling mode");
3657 goto failed_mount_wq;
3658 }
3659 default:
3660 break;
3661 }
3662 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3663
3664 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3665
3666 /*
3667 * The journal may have updated the bg summary counts, so we
3668 * need to update the global counters.
3669 */
3670 percpu_counter_set(&sbi->s_freeclusters_counter,
3671 ext4_count_free_clusters(sb));
3672 percpu_counter_set(&sbi->s_freeinodes_counter,
3673 ext4_count_free_inodes(sb));
3674 percpu_counter_set(&sbi->s_dirs_counter,
3675 ext4_count_dirs(sb));
3676 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3677
3678 no_journal:
3679 /*
3680 * Get the # of file system overhead blocks from the
3681 * superblock if present.
3682 */
3683 if (es->s_overhead_clusters)
3684 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3685 else {
3686 ret = ext4_calculate_overhead(sb);
3687 if (ret)
3688 goto failed_mount_wq;
3689 }
3690
3691 /*
3692 * The maximum number of concurrent works can be high and
3693 * concurrency isn't really necessary. Limit it to 1.
3694 */
3695 EXT4_SB(sb)->dio_unwritten_wq =
3696 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3697 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3698 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3699 goto failed_mount_wq;
3700 }
3701
3702 /*
3703 * The jbd2_journal_load will have done any necessary log recovery,
3704 * so we can safely mount the rest of the filesystem now.
3705 */
3706
3707 root = ext4_iget(sb, EXT4_ROOT_INO);
3708 if (IS_ERR(root)) {
3709 ext4_msg(sb, KERN_ERR, "get root inode failed");
3710 ret = PTR_ERR(root);
3711 root = NULL;
3712 goto failed_mount4;
3713 }
3714 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3715 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3716 iput(root);
3717 goto failed_mount4;
3718 }
3719 sb->s_root = d_make_root(root);
3720 if (!sb->s_root) {
3721 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3722 ret = -ENOMEM;
3723 goto failed_mount4;
3724 }
3725
3726 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3727 sb->s_flags |= MS_RDONLY;
3728
3729 /* determine the minimum size of new large inodes, if present */
3730 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3731 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3732 EXT4_GOOD_OLD_INODE_SIZE;
3733 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3734 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3735 if (sbi->s_want_extra_isize <
3736 le16_to_cpu(es->s_want_extra_isize))
3737 sbi->s_want_extra_isize =
3738 le16_to_cpu(es->s_want_extra_isize);
3739 if (sbi->s_want_extra_isize <
3740 le16_to_cpu(es->s_min_extra_isize))
3741 sbi->s_want_extra_isize =
3742 le16_to_cpu(es->s_min_extra_isize);
3743 }
3744 }
3745 /* Check if enough inode space is available */
3746 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3747 sbi->s_inode_size) {
3748 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3749 EXT4_GOOD_OLD_INODE_SIZE;
3750 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3751 "available");
3752 }
3753
3754 err = ext4_setup_system_zone(sb);
3755 if (err) {
3756 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3757 "zone (%d)", err);
3758 goto failed_mount4a;
3759 }
3760
3761 ext4_ext_init(sb);
3762 err = ext4_mb_init(sb, needs_recovery);
3763 if (err) {
3764 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3765 err);
3766 goto failed_mount5;
3767 }
3768
3769 err = ext4_register_li_request(sb, first_not_zeroed);
3770 if (err)
3771 goto failed_mount6;
3772
3773 sbi->s_kobj.kset = ext4_kset;
3774 init_completion(&sbi->s_kobj_unregister);
3775 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3776 "%s", sb->s_id);
3777 if (err)
3778 goto failed_mount7;
3779
3780 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3781 ext4_orphan_cleanup(sb, es);
3782 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3783 if (needs_recovery) {
3784 ext4_msg(sb, KERN_INFO, "recovery complete");
3785 ext4_mark_recovery_complete(sb, es);
3786 }
3787 if (EXT4_SB(sb)->s_journal) {
3788 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3789 descr = " journalled data mode";
3790 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3791 descr = " ordered data mode";
3792 else
3793 descr = " writeback data mode";
3794 } else
3795 descr = "out journal";
3796
3797 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3798 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3799 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3800
3801 if (es->s_error_count)
3802 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3803
3804 kfree(orig_data);
3805 return 0;
3806
3807 cantfind_ext4:
3808 if (!silent)
3809 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3810 goto failed_mount;
3811
3812 failed_mount7:
3813 ext4_unregister_li_request(sb);
3814 failed_mount6:
3815 ext4_mb_release(sb);
3816 failed_mount5:
3817 ext4_ext_release(sb);
3818 ext4_release_system_zone(sb);
3819 failed_mount4a:
3820 dput(sb->s_root);
3821 sb->s_root = NULL;
3822 failed_mount4:
3823 ext4_msg(sb, KERN_ERR, "mount failed");
3824 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3825 failed_mount_wq:
3826 if (sbi->s_journal) {
3827 jbd2_journal_destroy(sbi->s_journal);
3828 sbi->s_journal = NULL;
3829 }
3830 failed_mount3:
3831 del_timer(&sbi->s_err_report);
3832 if (sbi->s_flex_groups)
3833 ext4_kvfree(sbi->s_flex_groups);
3834 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3835 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3836 percpu_counter_destroy(&sbi->s_dirs_counter);
3837 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3838 if (sbi->s_mmp_tsk)
3839 kthread_stop(sbi->s_mmp_tsk);
3840 failed_mount2:
3841 for (i = 0; i < db_count; i++)
3842 brelse(sbi->s_group_desc[i]);
3843 ext4_kvfree(sbi->s_group_desc);
3844 failed_mount:
3845 if (sbi->s_proc) {
3846 remove_proc_entry("options", sbi->s_proc);
3847 remove_proc_entry(sb->s_id, ext4_proc_root);
3848 }
3849 #ifdef CONFIG_QUOTA
3850 for (i = 0; i < MAXQUOTAS; i++)
3851 kfree(sbi->s_qf_names[i]);
3852 #endif
3853 ext4_blkdev_remove(sbi);
3854 brelse(bh);
3855 out_fail:
3856 sb->s_fs_info = NULL;
3857 kfree(sbi->s_blockgroup_lock);
3858 kfree(sbi);
3859 out_free_orig:
3860 kfree(orig_data);
3861 return ret;
3862 }
3863
3864 /*
3865 * Setup any per-fs journal parameters now. We'll do this both on
3866 * initial mount, once the journal has been initialised but before we've
3867 * done any recovery; and again on any subsequent remount.
3868 */
ext4_init_journal_params(struct super_block * sb,journal_t * journal)3869 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3870 {
3871 struct ext4_sb_info *sbi = EXT4_SB(sb);
3872
3873 journal->j_commit_interval = sbi->s_commit_interval;
3874 journal->j_min_batch_time = sbi->s_min_batch_time;
3875 journal->j_max_batch_time = sbi->s_max_batch_time;
3876
3877 write_lock(&journal->j_state_lock);
3878 if (test_opt(sb, BARRIER))
3879 journal->j_flags |= JBD2_BARRIER;
3880 else
3881 journal->j_flags &= ~JBD2_BARRIER;
3882 if (test_opt(sb, DATA_ERR_ABORT))
3883 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3884 else
3885 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3886 write_unlock(&journal->j_state_lock);
3887 }
3888
ext4_get_journal(struct super_block * sb,unsigned int journal_inum)3889 static journal_t *ext4_get_journal(struct super_block *sb,
3890 unsigned int journal_inum)
3891 {
3892 struct inode *journal_inode;
3893 journal_t *journal;
3894
3895 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3896
3897 /* First, test for the existence of a valid inode on disk. Bad
3898 * things happen if we iget() an unused inode, as the subsequent
3899 * iput() will try to delete it. */
3900
3901 journal_inode = ext4_iget(sb, journal_inum);
3902 if (IS_ERR(journal_inode)) {
3903 ext4_msg(sb, KERN_ERR, "no journal found");
3904 return NULL;
3905 }
3906 if (!journal_inode->i_nlink) {
3907 make_bad_inode(journal_inode);
3908 iput(journal_inode);
3909 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3910 return NULL;
3911 }
3912
3913 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3914 journal_inode, journal_inode->i_size);
3915 if (!S_ISREG(journal_inode->i_mode)) {
3916 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3917 iput(journal_inode);
3918 return NULL;
3919 }
3920
3921 journal = jbd2_journal_init_inode(journal_inode);
3922 if (!journal) {
3923 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3924 iput(journal_inode);
3925 return NULL;
3926 }
3927 journal->j_private = sb;
3928 ext4_init_journal_params(sb, journal);
3929 return journal;
3930 }
3931
ext4_get_dev_journal(struct super_block * sb,dev_t j_dev)3932 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3933 dev_t j_dev)
3934 {
3935 struct buffer_head *bh;
3936 journal_t *journal;
3937 ext4_fsblk_t start;
3938 ext4_fsblk_t len;
3939 int hblock, blocksize;
3940 ext4_fsblk_t sb_block;
3941 unsigned long offset;
3942 struct ext4_super_block *es;
3943 struct block_device *bdev;
3944
3945 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3946
3947 bdev = ext4_blkdev_get(j_dev, sb);
3948 if (bdev == NULL)
3949 return NULL;
3950
3951 blocksize = sb->s_blocksize;
3952 hblock = bdev_logical_block_size(bdev);
3953 if (blocksize < hblock) {
3954 ext4_msg(sb, KERN_ERR,
3955 "blocksize too small for journal device");
3956 goto out_bdev;
3957 }
3958
3959 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3960 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3961 set_blocksize(bdev, blocksize);
3962 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3963 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3964 "external journal");
3965 goto out_bdev;
3966 }
3967
3968 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3969 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3970 !(le32_to_cpu(es->s_feature_incompat) &
3971 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3972 ext4_msg(sb, KERN_ERR, "external journal has "
3973 "bad superblock");
3974 brelse(bh);
3975 goto out_bdev;
3976 }
3977
3978 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3979 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3980 brelse(bh);
3981 goto out_bdev;
3982 }
3983
3984 len = ext4_blocks_count(es);
3985 start = sb_block + 1;
3986 brelse(bh); /* we're done with the superblock */
3987
3988 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3989 start, len, blocksize);
3990 if (!journal) {
3991 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3992 goto out_bdev;
3993 }
3994 journal->j_private = sb;
3995 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3996 wait_on_buffer(journal->j_sb_buffer);
3997 if (!buffer_uptodate(journal->j_sb_buffer)) {
3998 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3999 goto out_journal;
4000 }
4001 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4002 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4003 "user (unsupported) - %d",
4004 be32_to_cpu(journal->j_superblock->s_nr_users));
4005 goto out_journal;
4006 }
4007 EXT4_SB(sb)->journal_bdev = bdev;
4008 ext4_init_journal_params(sb, journal);
4009 return journal;
4010
4011 out_journal:
4012 jbd2_journal_destroy(journal);
4013 out_bdev:
4014 ext4_blkdev_put(bdev);
4015 return NULL;
4016 }
4017
ext4_load_journal(struct super_block * sb,struct ext4_super_block * es,unsigned long journal_devnum)4018 static int ext4_load_journal(struct super_block *sb,
4019 struct ext4_super_block *es,
4020 unsigned long journal_devnum)
4021 {
4022 journal_t *journal;
4023 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4024 dev_t journal_dev;
4025 int err = 0;
4026 int really_read_only;
4027
4028 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4029
4030 if (journal_devnum &&
4031 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4032 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4033 "numbers have changed");
4034 journal_dev = new_decode_dev(journal_devnum);
4035 } else
4036 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4037
4038 really_read_only = bdev_read_only(sb->s_bdev);
4039
4040 /*
4041 * Are we loading a blank journal or performing recovery after a
4042 * crash? For recovery, we need to check in advance whether we
4043 * can get read-write access to the device.
4044 */
4045 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4046 if (sb->s_flags & MS_RDONLY) {
4047 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4048 "required on readonly filesystem");
4049 if (really_read_only) {
4050 ext4_msg(sb, KERN_ERR, "write access "
4051 "unavailable, cannot proceed");
4052 return -EROFS;
4053 }
4054 ext4_msg(sb, KERN_INFO, "write access will "
4055 "be enabled during recovery");
4056 }
4057 }
4058
4059 if (journal_inum && journal_dev) {
4060 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4061 "and inode journals!");
4062 return -EINVAL;
4063 }
4064
4065 if (journal_inum) {
4066 if (!(journal = ext4_get_journal(sb, journal_inum)))
4067 return -EINVAL;
4068 } else {
4069 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4070 return -EINVAL;
4071 }
4072
4073 if (!(journal->j_flags & JBD2_BARRIER))
4074 ext4_msg(sb, KERN_INFO, "barriers disabled");
4075
4076 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4077 err = jbd2_journal_wipe(journal, !really_read_only);
4078 if (!err) {
4079 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4080 if (save)
4081 memcpy(save, ((char *) es) +
4082 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4083 err = jbd2_journal_load(journal);
4084 if (save)
4085 memcpy(((char *) es) + EXT4_S_ERR_START,
4086 save, EXT4_S_ERR_LEN);
4087 kfree(save);
4088 }
4089
4090 if (err) {
4091 ext4_msg(sb, KERN_ERR, "error loading journal");
4092 jbd2_journal_destroy(journal);
4093 return err;
4094 }
4095
4096 EXT4_SB(sb)->s_journal = journal;
4097 ext4_clear_journal_err(sb, es);
4098
4099 if (!really_read_only && journal_devnum &&
4100 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4101 es->s_journal_dev = cpu_to_le32(journal_devnum);
4102
4103 /* Make sure we flush the recovery flag to disk. */
4104 ext4_commit_super(sb, 1);
4105 }
4106
4107 return 0;
4108 }
4109
ext4_commit_super(struct super_block * sb,int sync)4110 static int ext4_commit_super(struct super_block *sb, int sync)
4111 {
4112 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4113 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4114 int error = 0;
4115
4116 if (!sbh || block_device_ejected(sb))
4117 return error;
4118 if (buffer_write_io_error(sbh)) {
4119 /*
4120 * Oh, dear. A previous attempt to write the
4121 * superblock failed. This could happen because the
4122 * USB device was yanked out. Or it could happen to
4123 * be a transient write error and maybe the block will
4124 * be remapped. Nothing we can do but to retry the
4125 * write and hope for the best.
4126 */
4127 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4128 "superblock detected");
4129 clear_buffer_write_io_error(sbh);
4130 set_buffer_uptodate(sbh);
4131 }
4132 /*
4133 * If the file system is mounted read-only, don't update the
4134 * superblock write time. This avoids updating the superblock
4135 * write time when we are mounting the root file system
4136 * read/only but we need to replay the journal; at that point,
4137 * for people who are east of GMT and who make their clock
4138 * tick in localtime for Windows bug-for-bug compatibility,
4139 * the clock is set in the future, and this will cause e2fsck
4140 * to complain and force a full file system check.
4141 */
4142 if (!(sb->s_flags & MS_RDONLY))
4143 es->s_wtime = cpu_to_le32(get_seconds());
4144 if (sb->s_bdev->bd_part)
4145 es->s_kbytes_written =
4146 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4147 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4148 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4149 else
4150 es->s_kbytes_written =
4151 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4152 ext4_free_blocks_count_set(es,
4153 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4154 &EXT4_SB(sb)->s_freeclusters_counter)));
4155 es->s_free_inodes_count =
4156 cpu_to_le32(percpu_counter_sum_positive(
4157 &EXT4_SB(sb)->s_freeinodes_counter));
4158 sb->s_dirt = 0;
4159 BUFFER_TRACE(sbh, "marking dirty");
4160 mark_buffer_dirty(sbh);
4161 if (sync) {
4162 error = sync_dirty_buffer(sbh);
4163 if (error)
4164 return error;
4165
4166 error = buffer_write_io_error(sbh);
4167 if (error) {
4168 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4169 "superblock");
4170 clear_buffer_write_io_error(sbh);
4171 set_buffer_uptodate(sbh);
4172 }
4173 }
4174 return error;
4175 }
4176
4177 /*
4178 * Have we just finished recovery? If so, and if we are mounting (or
4179 * remounting) the filesystem readonly, then we will end up with a
4180 * consistent fs on disk. Record that fact.
4181 */
ext4_mark_recovery_complete(struct super_block * sb,struct ext4_super_block * es)4182 static void ext4_mark_recovery_complete(struct super_block *sb,
4183 struct ext4_super_block *es)
4184 {
4185 journal_t *journal = EXT4_SB(sb)->s_journal;
4186
4187 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4188 BUG_ON(journal != NULL);
4189 return;
4190 }
4191 jbd2_journal_lock_updates(journal);
4192 if (jbd2_journal_flush(journal) < 0)
4193 goto out;
4194
4195 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4196 sb->s_flags & MS_RDONLY) {
4197 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4198 ext4_commit_super(sb, 1);
4199 }
4200
4201 out:
4202 jbd2_journal_unlock_updates(journal);
4203 }
4204
4205 /*
4206 * If we are mounting (or read-write remounting) a filesystem whose journal
4207 * has recorded an error from a previous lifetime, move that error to the
4208 * main filesystem now.
4209 */
ext4_clear_journal_err(struct super_block * sb,struct ext4_super_block * es)4210 static void ext4_clear_journal_err(struct super_block *sb,
4211 struct ext4_super_block *es)
4212 {
4213 journal_t *journal;
4214 int j_errno;
4215 const char *errstr;
4216
4217 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4218
4219 journal = EXT4_SB(sb)->s_journal;
4220
4221 /*
4222 * Now check for any error status which may have been recorded in the
4223 * journal by a prior ext4_error() or ext4_abort()
4224 */
4225
4226 j_errno = jbd2_journal_errno(journal);
4227 if (j_errno) {
4228 char nbuf[16];
4229
4230 errstr = ext4_decode_error(sb, j_errno, nbuf);
4231 ext4_warning(sb, "Filesystem error recorded "
4232 "from previous mount: %s", errstr);
4233 ext4_warning(sb, "Marking fs in need of filesystem check.");
4234
4235 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4236 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4237 ext4_commit_super(sb, 1);
4238
4239 jbd2_journal_clear_err(journal);
4240 jbd2_journal_update_sb_errno(journal);
4241 }
4242 }
4243
4244 /*
4245 * Force the running and committing transactions to commit,
4246 * and wait on the commit.
4247 */
ext4_force_commit(struct super_block * sb)4248 int ext4_force_commit(struct super_block *sb)
4249 {
4250 journal_t *journal;
4251 int ret = 0;
4252
4253 if (sb->s_flags & MS_RDONLY)
4254 return 0;
4255
4256 journal = EXT4_SB(sb)->s_journal;
4257 if (journal) {
4258 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4259 ret = ext4_journal_force_commit(journal);
4260 }
4261
4262 return ret;
4263 }
4264
ext4_write_super(struct super_block * sb)4265 static void ext4_write_super(struct super_block *sb)
4266 {
4267 lock_super(sb);
4268 ext4_commit_super(sb, 1);
4269 unlock_super(sb);
4270 }
4271
ext4_sync_fs(struct super_block * sb,int wait)4272 static int ext4_sync_fs(struct super_block *sb, int wait)
4273 {
4274 int ret = 0;
4275 tid_t target;
4276 struct ext4_sb_info *sbi = EXT4_SB(sb);
4277
4278 trace_ext4_sync_fs(sb, wait);
4279 flush_workqueue(sbi->dio_unwritten_wq);
4280 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4281 if (wait)
4282 jbd2_log_wait_commit(sbi->s_journal, target);
4283 }
4284 return ret;
4285 }
4286
4287 /*
4288 * LVM calls this function before a (read-only) snapshot is created. This
4289 * gives us a chance to flush the journal completely and mark the fs clean.
4290 *
4291 * Note that only this function cannot bring a filesystem to be in a clean
4292 * state independently, because ext4 prevents a new handle from being started
4293 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4294 * the upper layer.
4295 */
ext4_freeze(struct super_block * sb)4296 static int ext4_freeze(struct super_block *sb)
4297 {
4298 int error = 0;
4299 journal_t *journal;
4300
4301 if (sb->s_flags & MS_RDONLY)
4302 return 0;
4303
4304 journal = EXT4_SB(sb)->s_journal;
4305
4306 /* Now we set up the journal barrier. */
4307 jbd2_journal_lock_updates(journal);
4308
4309 /*
4310 * Don't clear the needs_recovery flag if we failed to flush
4311 * the journal.
4312 */
4313 error = jbd2_journal_flush(journal);
4314 if (error < 0)
4315 goto out;
4316
4317 /* Journal blocked and flushed, clear needs_recovery flag. */
4318 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4319 error = ext4_commit_super(sb, 1);
4320 out:
4321 /* we rely on s_frozen to stop further updates */
4322 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4323 return error;
4324 }
4325
4326 /*
4327 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4328 * flag here, even though the filesystem is not technically dirty yet.
4329 */
ext4_unfreeze(struct super_block * sb)4330 static int ext4_unfreeze(struct super_block *sb)
4331 {
4332 if (sb->s_flags & MS_RDONLY)
4333 return 0;
4334
4335 lock_super(sb);
4336 /* Reset the needs_recovery flag before the fs is unlocked. */
4337 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4338 ext4_commit_super(sb, 1);
4339 unlock_super(sb);
4340 return 0;
4341 }
4342
4343 /*
4344 * Structure to save mount options for ext4_remount's benefit
4345 */
4346 struct ext4_mount_options {
4347 unsigned long s_mount_opt;
4348 unsigned long s_mount_opt2;
4349 uid_t s_resuid;
4350 gid_t s_resgid;
4351 unsigned long s_commit_interval;
4352 u32 s_min_batch_time, s_max_batch_time;
4353 #ifdef CONFIG_QUOTA
4354 int s_jquota_fmt;
4355 char *s_qf_names[MAXQUOTAS];
4356 #endif
4357 };
4358
ext4_remount(struct super_block * sb,int * flags,char * data)4359 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4360 {
4361 struct ext4_super_block *es;
4362 struct ext4_sb_info *sbi = EXT4_SB(sb);
4363 unsigned long old_sb_flags;
4364 struct ext4_mount_options old_opts;
4365 int enable_quota = 0;
4366 ext4_group_t g;
4367 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4368 int err = 0;
4369 #ifdef CONFIG_QUOTA
4370 int i;
4371 #endif
4372 char *orig_data = kstrdup(data, GFP_KERNEL);
4373
4374 /* Store the original options */
4375 lock_super(sb);
4376 old_sb_flags = sb->s_flags;
4377 old_opts.s_mount_opt = sbi->s_mount_opt;
4378 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4379 old_opts.s_resuid = sbi->s_resuid;
4380 old_opts.s_resgid = sbi->s_resgid;
4381 old_opts.s_commit_interval = sbi->s_commit_interval;
4382 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4383 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4384 #ifdef CONFIG_QUOTA
4385 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4386 for (i = 0; i < MAXQUOTAS; i++)
4387 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4388 #endif
4389 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4390 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4391
4392 /*
4393 * Allow the "check" option to be passed as a remount option.
4394 */
4395 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4396 err = -EINVAL;
4397 goto restore_opts;
4398 }
4399
4400 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4401 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4402 ext4_msg(sb, KERN_ERR, "can't mount with "
4403 "both data=journal and delalloc");
4404 err = -EINVAL;
4405 goto restore_opts;
4406 }
4407 if (test_opt(sb, DIOREAD_NOLOCK)) {
4408 ext4_msg(sb, KERN_ERR, "can't mount with "
4409 "both data=journal and dioread_nolock");
4410 err = -EINVAL;
4411 goto restore_opts;
4412 }
4413 }
4414
4415 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4416 ext4_abort(sb, "Abort forced by user");
4417
4418 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4419 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4420
4421 es = sbi->s_es;
4422
4423 if (sbi->s_journal) {
4424 ext4_init_journal_params(sb, sbi->s_journal);
4425 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4426 }
4427
4428 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4429 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4430 err = -EROFS;
4431 goto restore_opts;
4432 }
4433
4434 if (*flags & MS_RDONLY) {
4435 err = dquot_suspend(sb, -1);
4436 if (err < 0)
4437 goto restore_opts;
4438
4439 /*
4440 * First of all, the unconditional stuff we have to do
4441 * to disable replay of the journal when we next remount
4442 */
4443 sb->s_flags |= MS_RDONLY;
4444
4445 /*
4446 * OK, test if we are remounting a valid rw partition
4447 * readonly, and if so set the rdonly flag and then
4448 * mark the partition as valid again.
4449 */
4450 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4451 (sbi->s_mount_state & EXT4_VALID_FS))
4452 es->s_state = cpu_to_le16(sbi->s_mount_state);
4453
4454 if (sbi->s_journal)
4455 ext4_mark_recovery_complete(sb, es);
4456 } else {
4457 /* Make sure we can mount this feature set readwrite */
4458 if (!ext4_feature_set_ok(sb, 0)) {
4459 err = -EROFS;
4460 goto restore_opts;
4461 }
4462 /*
4463 * Make sure the group descriptor checksums
4464 * are sane. If they aren't, refuse to remount r/w.
4465 */
4466 for (g = 0; g < sbi->s_groups_count; g++) {
4467 struct ext4_group_desc *gdp =
4468 ext4_get_group_desc(sb, g, NULL);
4469
4470 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4471 ext4_msg(sb, KERN_ERR,
4472 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4473 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4474 le16_to_cpu(gdp->bg_checksum));
4475 err = -EINVAL;
4476 goto restore_opts;
4477 }
4478 }
4479
4480 /*
4481 * If we have an unprocessed orphan list hanging
4482 * around from a previously readonly bdev mount,
4483 * require a full umount/remount for now.
4484 */
4485 if (es->s_last_orphan) {
4486 ext4_msg(sb, KERN_WARNING, "Couldn't "
4487 "remount RDWR because of unprocessed "
4488 "orphan inode list. Please "
4489 "umount/remount instead");
4490 err = -EINVAL;
4491 goto restore_opts;
4492 }
4493
4494 /*
4495 * Mounting a RDONLY partition read-write, so reread
4496 * and store the current valid flag. (It may have
4497 * been changed by e2fsck since we originally mounted
4498 * the partition.)
4499 */
4500 if (sbi->s_journal)
4501 ext4_clear_journal_err(sb, es);
4502 sbi->s_mount_state = le16_to_cpu(es->s_state);
4503 if (!ext4_setup_super(sb, es, 0))
4504 sb->s_flags &= ~MS_RDONLY;
4505 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4506 EXT4_FEATURE_INCOMPAT_MMP))
4507 if (ext4_multi_mount_protect(sb,
4508 le64_to_cpu(es->s_mmp_block))) {
4509 err = -EROFS;
4510 goto restore_opts;
4511 }
4512 enable_quota = 1;
4513 }
4514 }
4515
4516 /*
4517 * Reinitialize lazy itable initialization thread based on
4518 * current settings
4519 */
4520 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4521 ext4_unregister_li_request(sb);
4522 else {
4523 ext4_group_t first_not_zeroed;
4524 first_not_zeroed = ext4_has_uninit_itable(sb);
4525 ext4_register_li_request(sb, first_not_zeroed);
4526 }
4527
4528 ext4_setup_system_zone(sb);
4529 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4530 ext4_commit_super(sb, 1);
4531
4532 #ifdef CONFIG_QUOTA
4533 /* Release old quota file names */
4534 for (i = 0; i < MAXQUOTAS; i++)
4535 if (old_opts.s_qf_names[i] &&
4536 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4537 kfree(old_opts.s_qf_names[i]);
4538 #endif
4539 unlock_super(sb);
4540 if (enable_quota)
4541 dquot_resume(sb, -1);
4542
4543 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4544 kfree(orig_data);
4545 return 0;
4546
4547 restore_opts:
4548 sb->s_flags = old_sb_flags;
4549 sbi->s_mount_opt = old_opts.s_mount_opt;
4550 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4551 sbi->s_resuid = old_opts.s_resuid;
4552 sbi->s_resgid = old_opts.s_resgid;
4553 sbi->s_commit_interval = old_opts.s_commit_interval;
4554 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4555 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4556 #ifdef CONFIG_QUOTA
4557 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4558 for (i = 0; i < MAXQUOTAS; i++) {
4559 if (sbi->s_qf_names[i] &&
4560 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4561 kfree(sbi->s_qf_names[i]);
4562 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4563 }
4564 #endif
4565 unlock_super(sb);
4566 kfree(orig_data);
4567 return err;
4568 }
4569
ext4_statfs(struct dentry * dentry,struct kstatfs * buf)4570 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4571 {
4572 struct super_block *sb = dentry->d_sb;
4573 struct ext4_sb_info *sbi = EXT4_SB(sb);
4574 struct ext4_super_block *es = sbi->s_es;
4575 ext4_fsblk_t overhead = 0;
4576 u64 fsid;
4577 s64 bfree;
4578
4579 if (!test_opt(sb, MINIX_DF))
4580 overhead = sbi->s_overhead;
4581
4582 buf->f_type = EXT4_SUPER_MAGIC;
4583 buf->f_bsize = sb->s_blocksize;
4584 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, sbi->s_overhead);
4585 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4586 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4587 /* prevent underflow in case that few free space is available */
4588 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4589 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4590 if (buf->f_bfree < ext4_r_blocks_count(es))
4591 buf->f_bavail = 0;
4592 buf->f_files = le32_to_cpu(es->s_inodes_count);
4593 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4594 buf->f_namelen = EXT4_NAME_LEN;
4595 fsid = le64_to_cpup((void *)es->s_uuid) ^
4596 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4597 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4598 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4599
4600 return 0;
4601 }
4602
4603 /* Helper function for writing quotas on sync - we need to start transaction
4604 * before quota file is locked for write. Otherwise the are possible deadlocks:
4605 * Process 1 Process 2
4606 * ext4_create() quota_sync()
4607 * jbd2_journal_start() write_dquot()
4608 * dquot_initialize() down(dqio_mutex)
4609 * down(dqio_mutex) jbd2_journal_start()
4610 *
4611 */
4612
4613 #ifdef CONFIG_QUOTA
4614
dquot_to_inode(struct dquot * dquot)4615 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4616 {
4617 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4618 }
4619
ext4_write_dquot(struct dquot * dquot)4620 static int ext4_write_dquot(struct dquot *dquot)
4621 {
4622 int ret, err;
4623 handle_t *handle;
4624 struct inode *inode;
4625
4626 inode = dquot_to_inode(dquot);
4627 handle = ext4_journal_start(inode,
4628 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4629 if (IS_ERR(handle))
4630 return PTR_ERR(handle);
4631 ret = dquot_commit(dquot);
4632 err = ext4_journal_stop(handle);
4633 if (!ret)
4634 ret = err;
4635 return ret;
4636 }
4637
ext4_acquire_dquot(struct dquot * dquot)4638 static int ext4_acquire_dquot(struct dquot *dquot)
4639 {
4640 int ret, err;
4641 handle_t *handle;
4642
4643 handle = ext4_journal_start(dquot_to_inode(dquot),
4644 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4645 if (IS_ERR(handle))
4646 return PTR_ERR(handle);
4647 ret = dquot_acquire(dquot);
4648 err = ext4_journal_stop(handle);
4649 if (!ret)
4650 ret = err;
4651 return ret;
4652 }
4653
ext4_release_dquot(struct dquot * dquot)4654 static int ext4_release_dquot(struct dquot *dquot)
4655 {
4656 int ret, err;
4657 handle_t *handle;
4658
4659 handle = ext4_journal_start(dquot_to_inode(dquot),
4660 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4661 if (IS_ERR(handle)) {
4662 /* Release dquot anyway to avoid endless cycle in dqput() */
4663 dquot_release(dquot);
4664 return PTR_ERR(handle);
4665 }
4666 ret = dquot_release(dquot);
4667 err = ext4_journal_stop(handle);
4668 if (!ret)
4669 ret = err;
4670 return ret;
4671 }
4672
ext4_mark_dquot_dirty(struct dquot * dquot)4673 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4674 {
4675 /* Are we journaling quotas? */
4676 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4677 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4678 dquot_mark_dquot_dirty(dquot);
4679 return ext4_write_dquot(dquot);
4680 } else {
4681 return dquot_mark_dquot_dirty(dquot);
4682 }
4683 }
4684
ext4_write_info(struct super_block * sb,int type)4685 static int ext4_write_info(struct super_block *sb, int type)
4686 {
4687 int ret, err;
4688 handle_t *handle;
4689
4690 /* Data block + inode block */
4691 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4692 if (IS_ERR(handle))
4693 return PTR_ERR(handle);
4694 ret = dquot_commit_info(sb, type);
4695 err = ext4_journal_stop(handle);
4696 if (!ret)
4697 ret = err;
4698 return ret;
4699 }
4700
4701 /*
4702 * Turn on quotas during mount time - we need to find
4703 * the quota file and such...
4704 */
ext4_quota_on_mount(struct super_block * sb,int type)4705 static int ext4_quota_on_mount(struct super_block *sb, int type)
4706 {
4707 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4708 EXT4_SB(sb)->s_jquota_fmt, type);
4709 }
4710
4711 /*
4712 * Standard function to be called on quota_on
4713 */
ext4_quota_on(struct super_block * sb,int type,int format_id,struct path * path)4714 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4715 struct path *path)
4716 {
4717 int err;
4718
4719 if (!test_opt(sb, QUOTA))
4720 return -EINVAL;
4721
4722 /* Quotafile not on the same filesystem? */
4723 if (path->dentry->d_sb != sb)
4724 return -EXDEV;
4725 /* Journaling quota? */
4726 if (EXT4_SB(sb)->s_qf_names[type]) {
4727 /* Quotafile not in fs root? */
4728 if (path->dentry->d_parent != sb->s_root)
4729 ext4_msg(sb, KERN_WARNING,
4730 "Quota file not on filesystem root. "
4731 "Journaled quota will not work");
4732 }
4733
4734 /*
4735 * When we journal data on quota file, we have to flush journal to see
4736 * all updates to the file when we bypass pagecache...
4737 */
4738 if (EXT4_SB(sb)->s_journal &&
4739 ext4_should_journal_data(path->dentry->d_inode)) {
4740 /*
4741 * We don't need to lock updates but journal_flush() could
4742 * otherwise be livelocked...
4743 */
4744 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4745 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4746 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4747 if (err)
4748 return err;
4749 }
4750
4751 return dquot_quota_on(sb, type, format_id, path);
4752 }
4753
ext4_quota_off(struct super_block * sb,int type)4754 static int ext4_quota_off(struct super_block *sb, int type)
4755 {
4756 struct inode *inode = sb_dqopt(sb)->files[type];
4757 handle_t *handle;
4758
4759 /* Force all delayed allocation blocks to be allocated.
4760 * Caller already holds s_umount sem */
4761 if (test_opt(sb, DELALLOC))
4762 sync_filesystem(sb);
4763
4764 if (!inode)
4765 goto out;
4766
4767 /* Update modification times of quota files when userspace can
4768 * start looking at them */
4769 handle = ext4_journal_start(inode, 1);
4770 if (IS_ERR(handle))
4771 goto out;
4772 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4773 ext4_mark_inode_dirty(handle, inode);
4774 ext4_journal_stop(handle);
4775
4776 out:
4777 return dquot_quota_off(sb, type);
4778 }
4779
4780 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4781 * acquiring the locks... As quota files are never truncated and quota code
4782 * itself serializes the operations (and no one else should touch the files)
4783 * we don't have to be afraid of races */
ext4_quota_read(struct super_block * sb,int type,char * data,size_t len,loff_t off)4784 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4785 size_t len, loff_t off)
4786 {
4787 struct inode *inode = sb_dqopt(sb)->files[type];
4788 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4789 int err = 0;
4790 int offset = off & (sb->s_blocksize - 1);
4791 int tocopy;
4792 size_t toread;
4793 struct buffer_head *bh;
4794 loff_t i_size = i_size_read(inode);
4795
4796 if (off > i_size)
4797 return 0;
4798 if (off+len > i_size)
4799 len = i_size-off;
4800 toread = len;
4801 while (toread > 0) {
4802 tocopy = sb->s_blocksize - offset < toread ?
4803 sb->s_blocksize - offset : toread;
4804 bh = ext4_bread(NULL, inode, blk, 0, &err);
4805 if (err)
4806 return err;
4807 if (!bh) /* A hole? */
4808 memset(data, 0, tocopy);
4809 else
4810 memcpy(data, bh->b_data+offset, tocopy);
4811 brelse(bh);
4812 offset = 0;
4813 toread -= tocopy;
4814 data += tocopy;
4815 blk++;
4816 }
4817 return len;
4818 }
4819
4820 /* Write to quotafile (we know the transaction is already started and has
4821 * enough credits) */
ext4_quota_write(struct super_block * sb,int type,const char * data,size_t len,loff_t off)4822 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4823 const char *data, size_t len, loff_t off)
4824 {
4825 struct inode *inode = sb_dqopt(sb)->files[type];
4826 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4827 int err = 0;
4828 int offset = off & (sb->s_blocksize - 1);
4829 struct buffer_head *bh;
4830 handle_t *handle = journal_current_handle();
4831
4832 if (EXT4_SB(sb)->s_journal && !handle) {
4833 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4834 " cancelled because transaction is not started",
4835 (unsigned long long)off, (unsigned long long)len);
4836 return -EIO;
4837 }
4838 /*
4839 * Since we account only one data block in transaction credits,
4840 * then it is impossible to cross a block boundary.
4841 */
4842 if (sb->s_blocksize - offset < len) {
4843 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4844 " cancelled because not block aligned",
4845 (unsigned long long)off, (unsigned long long)len);
4846 return -EIO;
4847 }
4848
4849 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4850 bh = ext4_bread(handle, inode, blk, 1, &err);
4851 if (!bh)
4852 goto out;
4853 err = ext4_journal_get_write_access(handle, bh);
4854 if (err) {
4855 brelse(bh);
4856 goto out;
4857 }
4858 lock_buffer(bh);
4859 memcpy(bh->b_data+offset, data, len);
4860 flush_dcache_page(bh->b_page);
4861 unlock_buffer(bh);
4862 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4863 brelse(bh);
4864 out:
4865 if (err) {
4866 mutex_unlock(&inode->i_mutex);
4867 return err;
4868 }
4869 if (inode->i_size < off + len) {
4870 i_size_write(inode, off + len);
4871 EXT4_I(inode)->i_disksize = inode->i_size;
4872 ext4_mark_inode_dirty(handle, inode);
4873 }
4874 mutex_unlock(&inode->i_mutex);
4875 return len;
4876 }
4877
4878 #endif
4879
ext4_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)4880 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4881 const char *dev_name, void *data)
4882 {
4883 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4884 }
4885
4886 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
register_as_ext2(void)4887 static inline void register_as_ext2(void)
4888 {
4889 int err = register_filesystem(&ext2_fs_type);
4890 if (err)
4891 printk(KERN_WARNING
4892 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4893 }
4894
unregister_as_ext2(void)4895 static inline void unregister_as_ext2(void)
4896 {
4897 unregister_filesystem(&ext2_fs_type);
4898 }
4899
ext2_feature_set_ok(struct super_block * sb)4900 static inline int ext2_feature_set_ok(struct super_block *sb)
4901 {
4902 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4903 return 0;
4904 if (sb->s_flags & MS_RDONLY)
4905 return 1;
4906 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4907 return 0;
4908 return 1;
4909 }
4910 MODULE_ALIAS("ext2");
4911 #else
register_as_ext2(void)4912 static inline void register_as_ext2(void) { }
unregister_as_ext2(void)4913 static inline void unregister_as_ext2(void) { }
ext2_feature_set_ok(struct super_block * sb)4914 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
4915 #endif
4916
4917 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
register_as_ext3(void)4918 static inline void register_as_ext3(void)
4919 {
4920 int err = register_filesystem(&ext3_fs_type);
4921 if (err)
4922 printk(KERN_WARNING
4923 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4924 }
4925
unregister_as_ext3(void)4926 static inline void unregister_as_ext3(void)
4927 {
4928 unregister_filesystem(&ext3_fs_type);
4929 }
4930
ext3_feature_set_ok(struct super_block * sb)4931 static inline int ext3_feature_set_ok(struct super_block *sb)
4932 {
4933 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4934 return 0;
4935 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4936 return 0;
4937 if (sb->s_flags & MS_RDONLY)
4938 return 1;
4939 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4940 return 0;
4941 return 1;
4942 }
4943 MODULE_ALIAS("ext3");
4944 #else
register_as_ext3(void)4945 static inline void register_as_ext3(void) { }
unregister_as_ext3(void)4946 static inline void unregister_as_ext3(void) { }
ext3_feature_set_ok(struct super_block * sb)4947 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
4948 #endif
4949
4950 static struct file_system_type ext4_fs_type = {
4951 .owner = THIS_MODULE,
4952 .name = "ext4",
4953 .mount = ext4_mount,
4954 .kill_sb = kill_block_super,
4955 .fs_flags = FS_REQUIRES_DEV,
4956 };
4957
ext4_init_feat_adverts(void)4958 static int __init ext4_init_feat_adverts(void)
4959 {
4960 struct ext4_features *ef;
4961 int ret = -ENOMEM;
4962
4963 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4964 if (!ef)
4965 goto out;
4966
4967 ef->f_kobj.kset = ext4_kset;
4968 init_completion(&ef->f_kobj_unregister);
4969 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4970 "features");
4971 if (ret) {
4972 kfree(ef);
4973 goto out;
4974 }
4975
4976 ext4_feat = ef;
4977 ret = 0;
4978 out:
4979 return ret;
4980 }
4981
ext4_exit_feat_adverts(void)4982 static void ext4_exit_feat_adverts(void)
4983 {
4984 kobject_put(&ext4_feat->f_kobj);
4985 wait_for_completion(&ext4_feat->f_kobj_unregister);
4986 kfree(ext4_feat);
4987 }
4988
4989 /* Shared across all ext4 file systems */
4990 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4991 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4992
ext4_init_fs(void)4993 static int __init ext4_init_fs(void)
4994 {
4995 int i, err;
4996
4997 ext4_li_info = NULL;
4998 mutex_init(&ext4_li_mtx);
4999
5000 ext4_check_flag_values();
5001
5002 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5003 mutex_init(&ext4__aio_mutex[i]);
5004 init_waitqueue_head(&ext4__ioend_wq[i]);
5005 }
5006
5007 err = ext4_init_pageio();
5008 if (err)
5009 return err;
5010 err = ext4_init_system_zone();
5011 if (err)
5012 goto out6;
5013 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5014 if (!ext4_kset)
5015 goto out5;
5016 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5017
5018 err = ext4_init_feat_adverts();
5019 if (err)
5020 goto out4;
5021
5022 err = ext4_init_mballoc();
5023 if (err)
5024 goto out3;
5025
5026 err = ext4_init_xattr();
5027 if (err)
5028 goto out2;
5029 err = init_inodecache();
5030 if (err)
5031 goto out1;
5032 register_as_ext3();
5033 register_as_ext2();
5034 err = register_filesystem(&ext4_fs_type);
5035 if (err)
5036 goto out;
5037
5038 return 0;
5039 out:
5040 unregister_as_ext2();
5041 unregister_as_ext3();
5042 destroy_inodecache();
5043 out1:
5044 ext4_exit_xattr();
5045 out2:
5046 ext4_exit_mballoc();
5047 out3:
5048 ext4_exit_feat_adverts();
5049 out4:
5050 if (ext4_proc_root)
5051 remove_proc_entry("fs/ext4", NULL);
5052 kset_unregister(ext4_kset);
5053 out5:
5054 ext4_exit_system_zone();
5055 out6:
5056 ext4_exit_pageio();
5057 return err;
5058 }
5059
ext4_exit_fs(void)5060 static void __exit ext4_exit_fs(void)
5061 {
5062 ext4_destroy_lazyinit_thread();
5063 unregister_as_ext2();
5064 unregister_as_ext3();
5065 unregister_filesystem(&ext4_fs_type);
5066 destroy_inodecache();
5067 ext4_exit_xattr();
5068 ext4_exit_mballoc();
5069 ext4_exit_feat_adverts();
5070 remove_proc_entry("fs/ext4", NULL);
5071 kset_unregister(ext4_kset);
5072 ext4_exit_system_zone();
5073 ext4_exit_pageio();
5074 }
5075
5076 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5077 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5078 MODULE_LICENSE("GPL");
5079 module_init(ext4_init_fs)
5080 module_exit(ext4_exit_fs)
5081