]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - fs/ext4/super.c
projects / mirror_ubuntu-bionic-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
ext4: avoid panic during forced reboot due to aborted journal
[mirror_ubuntu-bionic-kernel.git] / fs / ext4 / super.c
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/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.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/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/dax.h>
41 #include <linux/cleancache.h>
42 #include <linux/uaccess.h>
43 #include <linux/user_namespace.h>
44
45 #include <linux/kthread.h>
46 #include <linux/freezer.h>
47
48 #include "ext4.h"
49 #include "ext4_extents.h" /* Needed for trace points definition */
50 #include "ext4_jbd2.h"
51 #include "xattr.h"
52 #include "acl.h"
53 #include "mballoc.h"
54 #include "fsmap.h"
55
56 #define CREATE_TRACE_POINTS
57 #include <trace/events/ext4.h>
58
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ratelimit_state ext4_mount_msg_ratelimit;
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 int ext4_remount(struct super_block *sb, int *flags, char *data);
73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
74 static int ext4_unfreeze(struct super_block *sb);
75 static int ext4_freeze(struct super_block *sb);
76 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
78 static inline int ext2_feature_set_ok(struct super_block *sb);
79 static inline int ext3_feature_set_ok(struct super_block *sb);
80 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
81 static void ext4_destroy_lazyinit_thread(void);
82 static void ext4_unregister_li_request(struct super_block *sb);
83 static void ext4_clear_request_list(void);
84 static struct inode *ext4_get_journal_inode(struct super_block *sb,
85 unsigned int journal_inum);
86
87 /*
88 * Lock ordering
89 *
90 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
91 * i_mmap_rwsem (inode->i_mmap_rwsem)!
92 *
93 * page fault path:
94 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
95 * page lock -> i_data_sem (rw)
96 *
97 * buffered write path:
98 * sb_start_write -> i_mutex -> mmap_sem
99 * sb_start_write -> i_mutex -> transaction start -> page lock ->
100 * i_data_sem (rw)
101 *
102 * truncate:
103 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
104 * i_mmap_rwsem (w) -> page lock
105 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
106 * transaction start -> i_data_sem (rw)
107 *
108 * direct IO:
109 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
110 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
111 * transaction start -> i_data_sem (rw)
112 *
113 * writepages:
114 * transaction start -> page lock(s) -> i_data_sem (rw)
115 */
116
117 static bool userns_mounts = false;
118 module_param(userns_mounts, bool, 0644);
119 MODULE_PARM_DESC(userns_mounts, "Allow mounts from unprivileged user namespaces");
120
121 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
122 static struct file_system_type ext2_fs_type = {
123 .owner = THIS_MODULE,
124 .name = "ext2",
125 .mount = ext4_mount,
126 .kill_sb = kill_block_super,
127 .fs_flags = FS_REQUIRES_DEV | FS_USERNS_MOUNT,
128 };
129 MODULE_ALIAS_FS("ext2");
130 MODULE_ALIAS("ext2");
131 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
132 #else
133 #define IS_EXT2_SB(sb) (0)
134 #endif
135
136
137 static struct file_system_type ext3_fs_type = {
138 .owner = THIS_MODULE,
139 .name = "ext3",
140 .mount = ext4_mount,
141 .kill_sb = kill_block_super,
142 .fs_flags = FS_REQUIRES_DEV | FS_USERNS_MOUNT,
143 };
144 MODULE_ALIAS_FS("ext3");
145 MODULE_ALIAS("ext3");
146 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
147
148 /*
149 * This works like sb_bread() except it uses ERR_PTR for error
150 * returns. Currently with sb_bread it's impossible to distinguish
151 * between ENOMEM and EIO situations (since both result in a NULL
152 * return.
153 */
154 struct buffer_head *
155 ext4_sb_bread(struct super_block *sb, sector_t block, int op_flags)
156 {
157 struct buffer_head *bh = sb_getblk(sb, block);
158
159 if (bh == NULL)
160 return ERR_PTR(-ENOMEM);
161 if (buffer_uptodate(bh))
162 return bh;
163 ll_rw_block(REQ_OP_READ, REQ_META | op_flags, 1, &bh);
164 wait_on_buffer(bh);
165 if (buffer_uptodate(bh))
166 return bh;
167 put_bh(bh);
168 return ERR_PTR(-EIO);
169 }
170
171 static int ext4_verify_csum_type(struct super_block *sb,
172 struct ext4_super_block *es)
173 {
174 if (!ext4_has_feature_metadata_csum(sb))
175 return 1;
176
177 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
178 }
179
180 static __le32 ext4_superblock_csum(struct super_block *sb,
181 struct ext4_super_block *es)
182 {
183 struct ext4_sb_info *sbi = EXT4_SB(sb);
184 int offset = offsetof(struct ext4_super_block, s_checksum);
185 __u32 csum;
186
187 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
188
189 return cpu_to_le32(csum);
190 }
191
192 static int ext4_superblock_csum_verify(struct super_block *sb,
193 struct ext4_super_block *es)
194 {
195 if (!ext4_has_metadata_csum(sb))
196 return 1;
197
198 return es->s_checksum == ext4_superblock_csum(sb, es);
199 }
200
201 void ext4_superblock_csum_set(struct super_block *sb)
202 {
203 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
204
205 if (!ext4_has_metadata_csum(sb))
206 return;
207
208 es->s_checksum = ext4_superblock_csum(sb, es);
209 }
210
211 void *ext4_kvmalloc(size_t size, gfp_t flags)
212 {
213 void *ret;
214
215 ret = kmalloc(size, flags | __GFP_NOWARN);
216 if (!ret)
217 ret = __vmalloc(size, flags, PAGE_KERNEL);
218 return ret;
219 }
220
221 void *ext4_kvzalloc(size_t size, gfp_t flags)
222 {
223 void *ret;
224
225 ret = kzalloc(size, flags | __GFP_NOWARN);
226 if (!ret)
227 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
228 return ret;
229 }
230
231 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
232 struct ext4_group_desc *bg)
233 {
234 return le32_to_cpu(bg->bg_block_bitmap_lo) |
235 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
236 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
237 }
238
239 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
240 struct ext4_group_desc *bg)
241 {
242 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
243 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
244 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
245 }
246
247 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
248 struct ext4_group_desc *bg)
249 {
250 return le32_to_cpu(bg->bg_inode_table_lo) |
251 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
252 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
253 }
254
255 __u32 ext4_free_group_clusters(struct super_block *sb,
256 struct ext4_group_desc *bg)
257 {
258 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
259 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
260 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
261 }
262
263 __u32 ext4_free_inodes_count(struct super_block *sb,
264 struct ext4_group_desc *bg)
265 {
266 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
267 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
268 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
269 }
270
271 __u32 ext4_used_dirs_count(struct super_block *sb,
272 struct ext4_group_desc *bg)
273 {
274 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
275 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
276 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
277 }
278
279 __u32 ext4_itable_unused_count(struct super_block *sb,
280 struct ext4_group_desc *bg)
281 {
282 return le16_to_cpu(bg->bg_itable_unused_lo) |
283 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
284 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
285 }
286
287 void ext4_block_bitmap_set(struct super_block *sb,
288 struct ext4_group_desc *bg, ext4_fsblk_t blk)
289 {
290 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
291 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
292 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
293 }
294
295 void ext4_inode_bitmap_set(struct super_block *sb,
296 struct ext4_group_desc *bg, ext4_fsblk_t blk)
297 {
298 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
299 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
300 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
301 }
302
303 void ext4_inode_table_set(struct super_block *sb,
304 struct ext4_group_desc *bg, ext4_fsblk_t blk)
305 {
306 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
307 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
308 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
309 }
310
311 void ext4_free_group_clusters_set(struct super_block *sb,
312 struct ext4_group_desc *bg, __u32 count)
313 {
314 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
315 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
316 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
317 }
318
319 void ext4_free_inodes_set(struct super_block *sb,
320 struct ext4_group_desc *bg, __u32 count)
321 {
322 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
323 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
324 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
325 }
326
327 void ext4_used_dirs_set(struct super_block *sb,
328 struct ext4_group_desc *bg, __u32 count)
329 {
330 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
331 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
332 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
333 }
334
335 void ext4_itable_unused_set(struct super_block *sb,
336 struct ext4_group_desc *bg, __u32 count)
337 {
338 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
339 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
340 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
341 }
342
343
344 static void __save_error_info(struct super_block *sb, const char *func,
345 unsigned int line)
346 {
347 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
348
349 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
350 if (bdev_read_only(sb->s_bdev))
351 return;
352 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
353 es->s_last_error_time = cpu_to_le32(get_seconds());
354 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
355 es->s_last_error_line = cpu_to_le32(line);
356 if (!es->s_first_error_time) {
357 es->s_first_error_time = es->s_last_error_time;
358 strncpy(es->s_first_error_func, func,
359 sizeof(es->s_first_error_func));
360 es->s_first_error_line = cpu_to_le32(line);
361 es->s_first_error_ino = es->s_last_error_ino;
362 es->s_first_error_block = es->s_last_error_block;
363 }
364 /*
365 * Start the daily error reporting function if it hasn't been
366 * started already
367 */
368 if (!es->s_error_count)
369 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
370 le32_add_cpu(&es->s_error_count, 1);
371 }
372
373 static void save_error_info(struct super_block *sb, const char *func,
374 unsigned int line)
375 {
376 __save_error_info(sb, func, line);
377 ext4_commit_super(sb, 1);
378 }
379
380 /*
381 * The del_gendisk() function uninitializes the disk-specific data
382 * structures, including the bdi structure, without telling anyone
383 * else. Once this happens, any attempt to call mark_buffer_dirty()
384 * (for example, by ext4_commit_super), will cause a kernel OOPS.
385 * This is a kludge to prevent these oops until we can put in a proper
386 * hook in del_gendisk() to inform the VFS and file system layers.
387 */
388 static int block_device_ejected(struct super_block *sb)
389 {
390 struct inode *bd_inode = sb->s_bdev->bd_inode;
391 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
392
393 return bdi->dev == NULL;
394 }
395
396 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
397 {
398 struct super_block *sb = journal->j_private;
399 struct ext4_sb_info *sbi = EXT4_SB(sb);
400 int error = is_journal_aborted(journal);
401 struct ext4_journal_cb_entry *jce;
402
403 BUG_ON(txn->t_state == T_FINISHED);
404
405 ext4_process_freed_data(sb, txn->t_tid);
406
407 spin_lock(&sbi->s_md_lock);
408 while (!list_empty(&txn->t_private_list)) {
409 jce = list_entry(txn->t_private_list.next,
410 struct ext4_journal_cb_entry, jce_list);
411 list_del_init(&jce->jce_list);
412 spin_unlock(&sbi->s_md_lock);
413 jce->jce_func(sb, jce, error);
414 spin_lock(&sbi->s_md_lock);
415 }
416 spin_unlock(&sbi->s_md_lock);
417 }
418
419 static bool system_going_down(void)
420 {
421 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
422 || system_state == SYSTEM_RESTART;
423 }
424
425 /* Deal with the reporting of failure conditions on a filesystem such as
426 * inconsistencies detected or read IO failures.
427 *
428 * On ext2, we can store the error state of the filesystem in the
429 * superblock. That is not possible on ext4, because we may have other
430 * write ordering constraints on the superblock which prevent us from
431 * writing it out straight away; and given that the journal is about to
432 * be aborted, we can't rely on the current, or future, transactions to
433 * write out the superblock safely.
434 *
435 * We'll just use the jbd2_journal_abort() error code to record an error in
436 * the journal instead. On recovery, the journal will complain about
437 * that error until we've noted it down and cleared it.
438 */
439
440 static void ext4_handle_error(struct super_block *sb)
441 {
442 if (sb_rdonly(sb))
443 return;
444
445 if (!test_opt(sb, ERRORS_CONT)) {
446 journal_t *journal = EXT4_SB(sb)->s_journal;
447
448 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
449 if (journal)
450 jbd2_journal_abort(journal, -EIO);
451 }
452 /*
453 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
454 * could panic during 'reboot -f' as the underlying device got already
455 * disabled.
456 */
457 if (test_opt(sb, ERRORS_RO) || system_going_down()) {
458 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
459 /*
460 * Make sure updated value of ->s_mount_flags will be visible
461 * before ->s_flags update
462 */
463 smp_wmb();
464 sb->s_flags |= SB_RDONLY;
465 } else if (test_opt(sb, ERRORS_PANIC)) {
466 if (EXT4_SB(sb)->s_journal &&
467 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
468 return;
469 panic("EXT4-fs (device %s): panic forced after error\n",
470 sb->s_id);
471 }
472 }
473
474 #define ext4_error_ratelimit(sb) \
475 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
476 "EXT4-fs error")
477
478 void __ext4_error(struct super_block *sb, const char *function,
479 unsigned int line, const char *fmt, ...)
480 {
481 struct va_format vaf;
482 va_list args;
483
484 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
485 return;
486
487 if (ext4_error_ratelimit(sb)) {
488 va_start(args, fmt);
489 vaf.fmt = fmt;
490 vaf.va = &args;
491 printk(KERN_CRIT
492 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
493 sb->s_id, function, line, current->comm, &vaf);
494 va_end(args);
495 }
496 save_error_info(sb, function, line);
497 ext4_handle_error(sb);
498 }
499
500 void __ext4_error_inode(struct inode *inode, const char *function,
501 unsigned int line, ext4_fsblk_t block,
502 const char *fmt, ...)
503 {
504 va_list args;
505 struct va_format vaf;
506 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
507
508 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
509 return;
510
511 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
512 es->s_last_error_block = cpu_to_le64(block);
513 if (ext4_error_ratelimit(inode->i_sb)) {
514 va_start(args, fmt);
515 vaf.fmt = fmt;
516 vaf.va = &args;
517 if (block)
518 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
519 "inode #%lu: block %llu: comm %s: %pV\n",
520 inode->i_sb->s_id, function, line, inode->i_ino,
521 block, current->comm, &vaf);
522 else
523 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
524 "inode #%lu: comm %s: %pV\n",
525 inode->i_sb->s_id, function, line, inode->i_ino,
526 current->comm, &vaf);
527 va_end(args);
528 }
529 save_error_info(inode->i_sb, function, line);
530 ext4_handle_error(inode->i_sb);
531 }
532
533 void __ext4_error_file(struct file *file, const char *function,
534 unsigned int line, ext4_fsblk_t block,
535 const char *fmt, ...)
536 {
537 va_list args;
538 struct va_format vaf;
539 struct ext4_super_block *es;
540 struct inode *inode = file_inode(file);
541 char pathname[80], *path;
542
543 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
544 return;
545
546 es = EXT4_SB(inode->i_sb)->s_es;
547 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
548 if (ext4_error_ratelimit(inode->i_sb)) {
549 path = file_path(file, pathname, sizeof(pathname));
550 if (IS_ERR(path))
551 path = "(unknown)";
552 va_start(args, fmt);
553 vaf.fmt = fmt;
554 vaf.va = &args;
555 if (block)
556 printk(KERN_CRIT
557 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
558 "block %llu: comm %s: path %s: %pV\n",
559 inode->i_sb->s_id, function, line, inode->i_ino,
560 block, current->comm, path, &vaf);
561 else
562 printk(KERN_CRIT
563 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
564 "comm %s: path %s: %pV\n",
565 inode->i_sb->s_id, function, line, inode->i_ino,
566 current->comm, path, &vaf);
567 va_end(args);
568 }
569 save_error_info(inode->i_sb, function, line);
570 ext4_handle_error(inode->i_sb);
571 }
572
573 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 -EFSCORRUPTED:
580 errstr = "Corrupt filesystem";
581 break;
582 case -EFSBADCRC:
583 errstr = "Filesystem failed CRC";
584 break;
585 case -EIO:
586 errstr = "IO failure";
587 break;
588 case -ENOMEM:
589 errstr = "Out of memory";
590 break;
591 case -EROFS:
592 if (!sb || (EXT4_SB(sb)->s_journal &&
593 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
594 errstr = "Journal has aborted";
595 else
596 errstr = "Readonly filesystem";
597 break;
598 default:
599 /* If the caller passed in an extra buffer for unknown
600 * errors, textualise them now. Else we just return
601 * NULL. */
602 if (nbuf) {
603 /* Check for truncated error codes... */
604 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
605 errstr = nbuf;
606 }
607 break;
608 }
609
610 return errstr;
611 }
612
613 /* __ext4_std_error decodes expected errors from journaling functions
614 * automatically and invokes the appropriate error response. */
615
616 void __ext4_std_error(struct super_block *sb, const char *function,
617 unsigned int line, int errno)
618 {
619 char nbuf[16];
620 const char *errstr;
621
622 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
623 return;
624
625 /* Special case: if the error is EROFS, and we're not already
626 * inside a transaction, then there's really no point in logging
627 * an error. */
628 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
629 return;
630
631 if (ext4_error_ratelimit(sb)) {
632 errstr = ext4_decode_error(sb, errno, nbuf);
633 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
634 sb->s_id, function, line, errstr);
635 }
636
637 save_error_info(sb, function, line);
638 ext4_handle_error(sb);
639 }
640
641 /*
642 * ext4_abort is a much stronger failure handler than ext4_error. The
643 * abort function may be used to deal with unrecoverable failures such
644 * as journal IO errors or ENOMEM at a critical moment in log management.
645 *
646 * We unconditionally force the filesystem into an ABORT|READONLY state,
647 * unless the error response on the fs has been set to panic in which
648 * case we take the easy way out and panic immediately.
649 */
650
651 void __ext4_abort(struct super_block *sb, const char *function,
652 unsigned int line, const char *fmt, ...)
653 {
654 struct va_format vaf;
655 va_list args;
656
657 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
658 return;
659
660 save_error_info(sb, function, line);
661 va_start(args, fmt);
662 vaf.fmt = fmt;
663 vaf.va = &args;
664 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
665 sb->s_id, function, line, &vaf);
666 va_end(args);
667
668 if (sb_rdonly(sb) == 0) {
669 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
670 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
671 /*
672 * Make sure updated value of ->s_mount_flags will be visible
673 * before ->s_flags update
674 */
675 smp_wmb();
676 sb->s_flags |= SB_RDONLY;
677 if (EXT4_SB(sb)->s_journal)
678 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
679 save_error_info(sb, function, line);
680 }
681 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
682 if (EXT4_SB(sb)->s_journal &&
683 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
684 return;
685 panic("EXT4-fs panic from previous error\n");
686 }
687 }
688
689 void __ext4_msg(struct super_block *sb,
690 const char *prefix, const char *fmt, ...)
691 {
692 struct va_format vaf;
693 va_list args;
694
695 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
696 return;
697
698 va_start(args, fmt);
699 vaf.fmt = fmt;
700 vaf.va = &args;
701 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
702 va_end(args);
703 }
704
705 #define ext4_warning_ratelimit(sb) \
706 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
707 "EXT4-fs warning")
708
709 void __ext4_warning(struct super_block *sb, const char *function,
710 unsigned int line, const char *fmt, ...)
711 {
712 struct va_format vaf;
713 va_list args;
714
715 if (!ext4_warning_ratelimit(sb))
716 return;
717
718 va_start(args, fmt);
719 vaf.fmt = fmt;
720 vaf.va = &args;
721 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
722 sb->s_id, function, line, &vaf);
723 va_end(args);
724 }
725
726 void __ext4_warning_inode(const struct inode *inode, const char *function,
727 unsigned int line, const char *fmt, ...)
728 {
729 struct va_format vaf;
730 va_list args;
731
732 if (!ext4_warning_ratelimit(inode->i_sb))
733 return;
734
735 va_start(args, fmt);
736 vaf.fmt = fmt;
737 vaf.va = &args;
738 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
739 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
740 function, line, inode->i_ino, current->comm, &vaf);
741 va_end(args);
742 }
743
744 void __ext4_grp_locked_error(const char *function, unsigned int line,
745 struct super_block *sb, ext4_group_t grp,
746 unsigned long ino, ext4_fsblk_t block,
747 const char *fmt, ...)
748 __releases(bitlock)
749 __acquires(bitlock)
750 {
751 struct va_format vaf;
752 va_list args;
753 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
754
755 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
756 return;
757
758 es->s_last_error_ino = cpu_to_le32(ino);
759 es->s_last_error_block = cpu_to_le64(block);
760 __save_error_info(sb, function, line);
761
762 if (ext4_error_ratelimit(sb)) {
763 va_start(args, fmt);
764 vaf.fmt = fmt;
765 vaf.va = &args;
766 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
767 sb->s_id, function, line, grp);
768 if (ino)
769 printk(KERN_CONT "inode %lu: ", ino);
770 if (block)
771 printk(KERN_CONT "block %llu:",
772 (unsigned long long) block);
773 printk(KERN_CONT "%pV\n", &vaf);
774 va_end(args);
775 }
776
777 if (test_opt(sb, ERRORS_CONT)) {
778 ext4_commit_super(sb, 0);
779 return;
780 }
781
782 ext4_unlock_group(sb, grp);
783 ext4_commit_super(sb, 1);
784 ext4_handle_error(sb);
785 /*
786 * We only get here in the ERRORS_RO case; relocking the group
787 * may be dangerous, but nothing bad will happen since the
788 * filesystem will have already been marked read/only and the
789 * journal has been aborted. We return 1 as a hint to callers
790 * who might what to use the return value from
791 * ext4_grp_locked_error() to distinguish between the
792 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
793 * aggressively from the ext4 function in question, with a
794 * more appropriate error code.
795 */
796 ext4_lock_group(sb, grp);
797 return;
798 }
799
800 void ext4_update_dynamic_rev(struct super_block *sb)
801 {
802 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
803
804 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
805 return;
806
807 ext4_warning(sb,
808 "updating to rev %d because of new feature flag, "
809 "running e2fsck is recommended",
810 EXT4_DYNAMIC_REV);
811
812 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
813 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
814 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
815 /* leave es->s_feature_*compat flags alone */
816 /* es->s_uuid will be set by e2fsck if empty */
817
818 /*
819 * The rest of the superblock fields should be zero, and if not it
820 * means they are likely already in use, so leave them alone. We
821 * can leave it up to e2fsck to clean up any inconsistencies there.
822 */
823 }
824
825 /*
826 * Open the external journal device
827 */
828 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
829 {
830 struct block_device *bdev;
831 char b[BDEVNAME_SIZE];
832
833 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
834 if (IS_ERR(bdev))
835 goto fail;
836 return bdev;
837
838 fail:
839 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
840 __bdevname(dev, b), PTR_ERR(bdev));
841 return NULL;
842 }
843
844 /*
845 * Release the journal device
846 */
847 static void ext4_blkdev_put(struct block_device *bdev)
848 {
849 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
850 }
851
852 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
853 {
854 struct block_device *bdev;
855 bdev = sbi->journal_bdev;
856 if (bdev) {
857 ext4_blkdev_put(bdev);
858 sbi->journal_bdev = NULL;
859 }
860 }
861
862 static inline struct inode *orphan_list_entry(struct list_head *l)
863 {
864 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
865 }
866
867 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
868 {
869 struct list_head *l;
870
871 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
872 le32_to_cpu(sbi->s_es->s_last_orphan));
873
874 printk(KERN_ERR "sb_info orphan list:\n");
875 list_for_each(l, &sbi->s_orphan) {
876 struct inode *inode = orphan_list_entry(l);
877 printk(KERN_ERR " "
878 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
879 inode->i_sb->s_id, inode->i_ino, inode,
880 inode->i_mode, inode->i_nlink,
881 NEXT_ORPHAN(inode));
882 }
883 }
884
885 #ifdef CONFIG_QUOTA
886 static int ext4_quota_off(struct super_block *sb, int type);
887
888 static inline void ext4_quota_off_umount(struct super_block *sb)
889 {
890 int type;
891
892 /* Use our quota_off function to clear inode flags etc. */
893 for (type = 0; type < EXT4_MAXQUOTAS; type++)
894 ext4_quota_off(sb, type);
895 }
896
897 /*
898 * This is a helper function which is used in the mount/remount
899 * codepaths (which holds s_umount) to fetch the quota file name.
900 */
901 static inline char *get_qf_name(struct super_block *sb,
902 struct ext4_sb_info *sbi,
903 int type)
904 {
905 return rcu_dereference_protected(sbi->s_qf_names[type],
906 lockdep_is_held(&sb->s_umount));
907 }
908 #else
909 static inline void ext4_quota_off_umount(struct super_block *sb)
910 {
911 }
912 #endif
913
914 static void ext4_put_super(struct super_block *sb)
915 {
916 struct ext4_sb_info *sbi = EXT4_SB(sb);
917 struct ext4_super_block *es = sbi->s_es;
918 int aborted = 0;
919 int i, err;
920
921 ext4_unregister_li_request(sb);
922 ext4_quota_off_umount(sb);
923
924 flush_workqueue(sbi->rsv_conversion_wq);
925 destroy_workqueue(sbi->rsv_conversion_wq);
926
927 if (sbi->s_journal) {
928 aborted = is_journal_aborted(sbi->s_journal);
929 err = jbd2_journal_destroy(sbi->s_journal);
930 sbi->s_journal = NULL;
931 if ((err < 0) && !aborted)
932 ext4_abort(sb, "Couldn't clean up the journal");
933 }
934
935 ext4_unregister_sysfs(sb);
936 ext4_es_unregister_shrinker(sbi);
937 del_timer_sync(&sbi->s_err_report);
938 ext4_release_system_zone(sb);
939 ext4_mb_release(sb);
940 ext4_ext_release(sb);
941
942 if (!sb_rdonly(sb) && !aborted) {
943 ext4_clear_feature_journal_needs_recovery(sb);
944 es->s_state = cpu_to_le16(sbi->s_mount_state);
945 }
946 if (!sb_rdonly(sb))
947 ext4_commit_super(sb, 1);
948
949 for (i = 0; i < sbi->s_gdb_count; i++)
950 brelse(sbi->s_group_desc[i]);
951 kvfree(sbi->s_group_desc);
952 kvfree(sbi->s_flex_groups);
953 percpu_counter_destroy(&sbi->s_freeclusters_counter);
954 percpu_counter_destroy(&sbi->s_freeinodes_counter);
955 percpu_counter_destroy(&sbi->s_dirs_counter);
956 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
957 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
958 #ifdef CONFIG_QUOTA
959 for (i = 0; i < EXT4_MAXQUOTAS; i++)
960 kfree(get_qf_name(sb, sbi, i));
961 #endif
962
963 /* Debugging code just in case the in-memory inode orphan list
964 * isn't empty. The on-disk one can be non-empty if we've
965 * detected an error and taken the fs readonly, but the
966 * in-memory list had better be clean by this point. */
967 if (!list_empty(&sbi->s_orphan))
968 dump_orphan_list(sb, sbi);
969 J_ASSERT(list_empty(&sbi->s_orphan));
970
971 sync_blockdev(sb->s_bdev);
972 invalidate_bdev(sb->s_bdev);
973 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
974 /*
975 * Invalidate the journal device's buffers. We don't want them
976 * floating about in memory - the physical journal device may
977 * hotswapped, and it breaks the `ro-after' testing code.
978 */
979 sync_blockdev(sbi->journal_bdev);
980 invalidate_bdev(sbi->journal_bdev);
981 ext4_blkdev_remove(sbi);
982 }
983 if (sbi->s_ea_inode_cache) {
984 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
985 sbi->s_ea_inode_cache = NULL;
986 }
987 if (sbi->s_ea_block_cache) {
988 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
989 sbi->s_ea_block_cache = NULL;
990 }
991 if (sbi->s_mmp_tsk)
992 kthread_stop(sbi->s_mmp_tsk);
993 brelse(sbi->s_sbh);
994 sb->s_fs_info = NULL;
995 /*
996 * Now that we are completely done shutting down the
997 * superblock, we need to actually destroy the kobject.
998 */
999 kobject_put(&sbi->s_kobj);
1000 wait_for_completion(&sbi->s_kobj_unregister);
1001 if (sbi->s_chksum_driver)
1002 crypto_free_shash(sbi->s_chksum_driver);
1003 kfree(sbi->s_blockgroup_lock);
1004 fs_put_dax(sbi->s_daxdev);
1005 kfree(sbi);
1006 }
1007
1008 static struct kmem_cache *ext4_inode_cachep;
1009
1010 /*
1011 * Called inside transaction, so use GFP_NOFS
1012 */
1013 static struct inode *ext4_alloc_inode(struct super_block *sb)
1014 {
1015 struct ext4_inode_info *ei;
1016
1017 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1018 if (!ei)
1019 return NULL;
1020
1021 ei->vfs_inode.i_version = 1;
1022 spin_lock_init(&ei->i_raw_lock);
1023 INIT_LIST_HEAD(&ei->i_prealloc_list);
1024 spin_lock_init(&ei->i_prealloc_lock);
1025 ext4_es_init_tree(&ei->i_es_tree);
1026 rwlock_init(&ei->i_es_lock);
1027 INIT_LIST_HEAD(&ei->i_es_list);
1028 ei->i_es_all_nr = 0;
1029 ei->i_es_shk_nr = 0;
1030 ei->i_es_shrink_lblk = 0;
1031 ei->i_reserved_data_blocks = 0;
1032 ei->i_da_metadata_calc_len = 0;
1033 ei->i_da_metadata_calc_last_lblock = 0;
1034 spin_lock_init(&(ei->i_block_reservation_lock));
1035 #ifdef CONFIG_QUOTA
1036 ei->i_reserved_quota = 0;
1037 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1038 #endif
1039 ei->jinode = NULL;
1040 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1041 spin_lock_init(&ei->i_completed_io_lock);
1042 ei->i_sync_tid = 0;
1043 ei->i_datasync_tid = 0;
1044 atomic_set(&ei->i_unwritten, 0);
1045 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1046 return &ei->vfs_inode;
1047 }
1048
1049 static int ext4_drop_inode(struct inode *inode)
1050 {
1051 int drop = generic_drop_inode(inode);
1052
1053 trace_ext4_drop_inode(inode, drop);
1054 return drop;
1055 }
1056
1057 static void ext4_i_callback(struct rcu_head *head)
1058 {
1059 struct inode *inode = container_of(head, struct inode, i_rcu);
1060 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1061 }
1062
1063 static void ext4_destroy_inode(struct inode *inode)
1064 {
1065 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1066 ext4_msg(inode->i_sb, KERN_ERR,
1067 "Inode %lu (%p): orphan list check failed!",
1068 inode->i_ino, EXT4_I(inode));
1069 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1070 EXT4_I(inode), sizeof(struct ext4_inode_info),
1071 true);
1072 dump_stack();
1073 }
1074 call_rcu(&inode->i_rcu, ext4_i_callback);
1075 }
1076
1077 static void init_once(void *foo)
1078 {
1079 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1080
1081 INIT_LIST_HEAD(&ei->i_orphan);
1082 init_rwsem(&ei->xattr_sem);
1083 init_rwsem(&ei->i_data_sem);
1084 init_rwsem(&ei->i_mmap_sem);
1085 inode_init_once(&ei->vfs_inode);
1086 }
1087
1088 static int __init init_inodecache(void)
1089 {
1090 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1091 sizeof(struct ext4_inode_info),
1092 0, (SLAB_RECLAIM_ACCOUNT|
1093 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1094 init_once);
1095 if (ext4_inode_cachep == NULL)
1096 return -ENOMEM;
1097 return 0;
1098 }
1099
1100 static void destroy_inodecache(void)
1101 {
1102 /*
1103 * Make sure all delayed rcu free inodes are flushed before we
1104 * destroy cache.
1105 */
1106 rcu_barrier();
1107 kmem_cache_destroy(ext4_inode_cachep);
1108 }
1109
1110 void ext4_clear_inode(struct inode *inode)
1111 {
1112 invalidate_inode_buffers(inode);
1113 clear_inode(inode);
1114 dquot_drop(inode);
1115 ext4_discard_preallocations(inode);
1116 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1117 if (EXT4_I(inode)->jinode) {
1118 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1119 EXT4_I(inode)->jinode);
1120 jbd2_free_inode(EXT4_I(inode)->jinode);
1121 EXT4_I(inode)->jinode = NULL;
1122 }
1123 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1124 fscrypt_put_encryption_info(inode, NULL);
1125 #endif
1126 }
1127
1128 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1129 u64 ino, u32 generation)
1130 {
1131 struct inode *inode;
1132
1133 /*
1134 * Currently we don't know the generation for parent directory, so
1135 * a generation of 0 means "accept any"
1136 */
1137 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1138 if (IS_ERR(inode))
1139 return ERR_CAST(inode);
1140 if (generation && inode->i_generation != generation) {
1141 iput(inode);
1142 return ERR_PTR(-ESTALE);
1143 }
1144
1145 return inode;
1146 }
1147
1148 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1149 int fh_len, int fh_type)
1150 {
1151 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1152 ext4_nfs_get_inode);
1153 }
1154
1155 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1156 int fh_len, int fh_type)
1157 {
1158 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1159 ext4_nfs_get_inode);
1160 }
1161
1162 static int ext4_nfs_commit_metadata(struct inode *inode)
1163 {
1164 struct writeback_control wbc = {
1165 .sync_mode = WB_SYNC_ALL
1166 };
1167
1168 trace_ext4_nfs_commit_metadata(inode);
1169 return ext4_write_inode(inode, &wbc);
1170 }
1171
1172 /*
1173 * Try to release metadata pages (indirect blocks, directories) which are
1174 * mapped via the block device. Since these pages could have journal heads
1175 * which would prevent try_to_free_buffers() from freeing them, we must use
1176 * jbd2 layer's try_to_free_buffers() function to release them.
1177 */
1178 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1179 gfp_t wait)
1180 {
1181 journal_t *journal = EXT4_SB(sb)->s_journal;
1182
1183 WARN_ON(PageChecked(page));
1184 if (!page_has_buffers(page))
1185 return 0;
1186 if (journal)
1187 return jbd2_journal_try_to_free_buffers(journal, page,
1188 wait & ~__GFP_DIRECT_RECLAIM);
1189 return try_to_free_buffers(page);
1190 }
1191
1192 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1193 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1194 {
1195 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1196 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1197 }
1198
1199 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1200 void *fs_data)
1201 {
1202 handle_t *handle = fs_data;
1203 int res, res2, credits, retries = 0;
1204
1205 /*
1206 * Encrypting the root directory is not allowed because e2fsck expects
1207 * lost+found to exist and be unencrypted, and encrypting the root
1208 * directory would imply encrypting the lost+found directory as well as
1209 * the filename "lost+found" itself.
1210 */
1211 if (inode->i_ino == EXT4_ROOT_INO)
1212 return -EPERM;
1213
1214 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1215 return -EINVAL;
1216
1217 res = ext4_convert_inline_data(inode);
1218 if (res)
1219 return res;
1220
1221 /*
1222 * If a journal handle was specified, then the encryption context is
1223 * being set on a new inode via inheritance and is part of a larger
1224 * transaction to create the inode. Otherwise the encryption context is
1225 * being set on an existing inode in its own transaction. Only in the
1226 * latter case should the "retry on ENOSPC" logic be used.
1227 */
1228
1229 if (handle) {
1230 res = ext4_xattr_set_handle(handle, inode,
1231 EXT4_XATTR_INDEX_ENCRYPTION,
1232 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1233 ctx, len, 0);
1234 if (!res) {
1235 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1236 ext4_clear_inode_state(inode,
1237 EXT4_STATE_MAY_INLINE_DATA);
1238 /*
1239 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1240 * S_DAX may be disabled
1241 */
1242 ext4_set_inode_flags(inode);
1243 }
1244 return res;
1245 }
1246
1247 res = dquot_initialize(inode);
1248 if (res)
1249 return res;
1250 retry:
1251 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1252 &credits);
1253 if (res)
1254 return res;
1255
1256 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1257 if (IS_ERR(handle))
1258 return PTR_ERR(handle);
1259
1260 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1261 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1262 ctx, len, 0);
1263 if (!res) {
1264 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1265 /*
1266 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1267 * S_DAX may be disabled
1268 */
1269 ext4_set_inode_flags(inode);
1270 res = ext4_mark_inode_dirty(handle, inode);
1271 if (res)
1272 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1273 }
1274 res2 = ext4_journal_stop(handle);
1275
1276 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1277 goto retry;
1278 if (!res)
1279 res = res2;
1280 return res;
1281 }
1282
1283 static bool ext4_dummy_context(struct inode *inode)
1284 {
1285 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1286 }
1287
1288 static unsigned ext4_max_namelen(struct inode *inode)
1289 {
1290 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1291 EXT4_NAME_LEN;
1292 }
1293
1294 static const struct fscrypt_operations ext4_cryptops = {
1295 .key_prefix = "ext4:",
1296 .get_context = ext4_get_context,
1297 .set_context = ext4_set_context,
1298 .dummy_context = ext4_dummy_context,
1299 .empty_dir = ext4_empty_dir,
1300 .max_namelen = ext4_max_namelen,
1301 };
1302 #endif
1303
1304 #ifdef CONFIG_QUOTA
1305 static const char * const quotatypes[] = INITQFNAMES;
1306 #define QTYPE2NAME(t) (quotatypes[t])
1307
1308 static int ext4_write_dquot(struct dquot *dquot);
1309 static int ext4_acquire_dquot(struct dquot *dquot);
1310 static int ext4_release_dquot(struct dquot *dquot);
1311 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1312 static int ext4_write_info(struct super_block *sb, int type);
1313 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1314 const struct path *path);
1315 static int ext4_quota_on_mount(struct super_block *sb, int type);
1316 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1317 size_t len, loff_t off);
1318 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1319 const char *data, size_t len, loff_t off);
1320 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1321 unsigned int flags);
1322 static int ext4_enable_quotas(struct super_block *sb);
1323 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1324
1325 static struct dquot **ext4_get_dquots(struct inode *inode)
1326 {
1327 return EXT4_I(inode)->i_dquot;
1328 }
1329
1330 static const struct dquot_operations ext4_quota_operations = {
1331 .get_reserved_space = ext4_get_reserved_space,
1332 .write_dquot = ext4_write_dquot,
1333 .acquire_dquot = ext4_acquire_dquot,
1334 .release_dquot = ext4_release_dquot,
1335 .mark_dirty = ext4_mark_dquot_dirty,
1336 .write_info = ext4_write_info,
1337 .alloc_dquot = dquot_alloc,
1338 .destroy_dquot = dquot_destroy,
1339 .get_projid = ext4_get_projid,
1340 .get_inode_usage = ext4_get_inode_usage,
1341 .get_next_id = ext4_get_next_id,
1342 };
1343
1344 static const struct quotactl_ops ext4_qctl_operations = {
1345 .quota_on = ext4_quota_on,
1346 .quota_off = ext4_quota_off,
1347 .quota_sync = dquot_quota_sync,
1348 .get_state = dquot_get_state,
1349 .set_info = dquot_set_dqinfo,
1350 .get_dqblk = dquot_get_dqblk,
1351 .set_dqblk = dquot_set_dqblk,
1352 .get_nextdqblk = dquot_get_next_dqblk,
1353 };
1354 #endif
1355
1356 static const struct super_operations ext4_sops = {
1357 .alloc_inode = ext4_alloc_inode,
1358 .destroy_inode = ext4_destroy_inode,
1359 .write_inode = ext4_write_inode,
1360 .dirty_inode = ext4_dirty_inode,
1361 .drop_inode = ext4_drop_inode,
1362 .evict_inode = ext4_evict_inode,
1363 .put_super = ext4_put_super,
1364 .sync_fs = ext4_sync_fs,
1365 .freeze_fs = ext4_freeze,
1366 .unfreeze_fs = ext4_unfreeze,
1367 .statfs = ext4_statfs,
1368 .remount_fs = ext4_remount,
1369 .show_options = ext4_show_options,
1370 #ifdef CONFIG_QUOTA
1371 .quota_read = ext4_quota_read,
1372 .quota_write = ext4_quota_write,
1373 .get_dquots = ext4_get_dquots,
1374 #endif
1375 .bdev_try_to_free_page = bdev_try_to_free_page,
1376 };
1377
1378 static const struct export_operations ext4_export_ops = {
1379 .fh_to_dentry = ext4_fh_to_dentry,
1380 .fh_to_parent = ext4_fh_to_parent,
1381 .get_parent = ext4_get_parent,
1382 .commit_metadata = ext4_nfs_commit_metadata,
1383 };
1384
1385 enum {
1386 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1387 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1388 Opt_nouid32, Opt_debug, Opt_removed,
1389 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1390 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1391 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1392 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1393 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1394 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1395 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1396 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1397 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1398 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1399 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1400 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1401 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1402 Opt_inode_readahead_blks, Opt_journal_ioprio,
1403 Opt_dioread_nolock, Opt_dioread_lock,
1404 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1405 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1406 };
1407
1408 static const match_table_t tokens = {
1409 {Opt_bsd_df, "bsddf"},
1410 {Opt_minix_df, "minixdf"},
1411 {Opt_grpid, "grpid"},
1412 {Opt_grpid, "bsdgroups"},
1413 {Opt_nogrpid, "nogrpid"},
1414 {Opt_nogrpid, "sysvgroups"},
1415 {Opt_resgid, "resgid=%u"},
1416 {Opt_resuid, "resuid=%u"},
1417 {Opt_sb, "sb=%u"},
1418 {Opt_err_cont, "errors=continue"},
1419 {Opt_err_panic, "errors=panic"},
1420 {Opt_err_ro, "errors=remount-ro"},
1421 {Opt_nouid32, "nouid32"},
1422 {Opt_debug, "debug"},
1423 {Opt_removed, "oldalloc"},
1424 {Opt_removed, "orlov"},
1425 {Opt_user_xattr, "user_xattr"},
1426 {Opt_nouser_xattr, "nouser_xattr"},
1427 {Opt_acl, "acl"},
1428 {Opt_noacl, "noacl"},
1429 {Opt_noload, "norecovery"},
1430 {Opt_noload, "noload"},
1431 {Opt_removed, "nobh"},
1432 {Opt_removed, "bh"},
1433 {Opt_commit, "commit=%u"},
1434 {Opt_min_batch_time, "min_batch_time=%u"},
1435 {Opt_max_batch_time, "max_batch_time=%u"},
1436 {Opt_journal_dev, "journal_dev=%u"},
1437 {Opt_journal_path, "journal_path=%s"},
1438 {Opt_journal_checksum, "journal_checksum"},
1439 {Opt_nojournal_checksum, "nojournal_checksum"},
1440 {Opt_journal_async_commit, "journal_async_commit"},
1441 {Opt_abort, "abort"},
1442 {Opt_data_journal, "data=journal"},
1443 {Opt_data_ordered, "data=ordered"},
1444 {Opt_data_writeback, "data=writeback"},
1445 {Opt_data_err_abort, "data_err=abort"},
1446 {Opt_data_err_ignore, "data_err=ignore"},
1447 {Opt_offusrjquota, "usrjquota="},
1448 {Opt_usrjquota, "usrjquota=%s"},
1449 {Opt_offgrpjquota, "grpjquota="},
1450 {Opt_grpjquota, "grpjquota=%s"},
1451 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1452 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1453 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1454 {Opt_grpquota, "grpquota"},
1455 {Opt_noquota, "noquota"},
1456 {Opt_quota, "quota"},
1457 {Opt_usrquota, "usrquota"},
1458 {Opt_prjquota, "prjquota"},
1459 {Opt_barrier, "barrier=%u"},
1460 {Opt_barrier, "barrier"},
1461 {Opt_nobarrier, "nobarrier"},
1462 {Opt_i_version, "i_version"},
1463 {Opt_dax, "dax"},
1464 {Opt_stripe, "stripe=%u"},
1465 {Opt_delalloc, "delalloc"},
1466 {Opt_lazytime, "lazytime"},
1467 {Opt_nolazytime, "nolazytime"},
1468 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1469 {Opt_nodelalloc, "nodelalloc"},
1470 {Opt_removed, "mblk_io_submit"},
1471 {Opt_removed, "nomblk_io_submit"},
1472 {Opt_block_validity, "block_validity"},
1473 {Opt_noblock_validity, "noblock_validity"},
1474 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1475 {Opt_journal_ioprio, "journal_ioprio=%u"},
1476 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1477 {Opt_auto_da_alloc, "auto_da_alloc"},
1478 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1479 {Opt_dioread_nolock, "dioread_nolock"},
1480 {Opt_dioread_lock, "dioread_lock"},
1481 {Opt_discard, "discard"},
1482 {Opt_nodiscard, "nodiscard"},
1483 {Opt_init_itable, "init_itable=%u"},
1484 {Opt_init_itable, "init_itable"},
1485 {Opt_noinit_itable, "noinit_itable"},
1486 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1487 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1488 {Opt_nombcache, "nombcache"},
1489 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1490 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1491 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1492 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1493 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1494 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1495 {Opt_err, NULL},
1496 };
1497
1498 static ext4_fsblk_t get_sb_block(void **data)
1499 {
1500 ext4_fsblk_t sb_block;
1501 char *options = (char *) *data;
1502
1503 if (!options || strncmp(options, "sb=", 3) != 0)
1504 return 1; /* Default location */
1505
1506 options += 3;
1507 /* TODO: use simple_strtoll with >32bit ext4 */
1508 sb_block = simple_strtoul(options, &options, 0);
1509 if (*options && *options != ',') {
1510 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1511 (char *) *data);
1512 return 1;
1513 }
1514 if (*options == ',')
1515 options++;
1516 *data = (void *) options;
1517
1518 return sb_block;
1519 }
1520
1521 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1522 static const char deprecated_msg[] =
1523 "Mount option \"%s\" will be removed by %s\n"
1524 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1525
1526 #ifdef CONFIG_QUOTA
1527 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1528 {
1529 struct ext4_sb_info *sbi = EXT4_SB(sb);
1530 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1531 int ret = -1;
1532
1533 if (sb_any_quota_loaded(sb) && !old_qname) {
1534 ext4_msg(sb, KERN_ERR,
1535 "Cannot change journaled "
1536 "quota options when quota turned on");
1537 return -1;
1538 }
1539 if (ext4_has_feature_quota(sb)) {
1540 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1541 "ignored when QUOTA feature is enabled");
1542 return 1;
1543 }
1544 qname = match_strdup(args);
1545 if (!qname) {
1546 ext4_msg(sb, KERN_ERR,
1547 "Not enough memory for storing quotafile name");
1548 return -1;
1549 }
1550 if (old_qname) {
1551 if (strcmp(old_qname, qname) == 0)
1552 ret = 1;
1553 else
1554 ext4_msg(sb, KERN_ERR,
1555 "%s quota file already specified",
1556 QTYPE2NAME(qtype));
1557 goto errout;
1558 }
1559 if (strchr(qname, '/')) {
1560 ext4_msg(sb, KERN_ERR,
1561 "quotafile must be on filesystem root");
1562 goto errout;
1563 }
1564 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1565 set_opt(sb, QUOTA);
1566 return 1;
1567 errout:
1568 kfree(qname);
1569 return ret;
1570 }
1571
1572 static int clear_qf_name(struct super_block *sb, int qtype)
1573 {
1574
1575 struct ext4_sb_info *sbi = EXT4_SB(sb);
1576 char *old_qname = get_qf_name(sb, sbi, qtype);
1577
1578 if (sb_any_quota_loaded(sb) && old_qname) {
1579 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1580 " when quota turned on");
1581 return -1;
1582 }
1583 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1584 synchronize_rcu();
1585 kfree(old_qname);
1586 return 1;
1587 }
1588 #endif
1589
1590 #define MOPT_SET 0x0001
1591 #define MOPT_CLEAR 0x0002
1592 #define MOPT_NOSUPPORT 0x0004
1593 #define MOPT_EXPLICIT 0x0008
1594 #define MOPT_CLEAR_ERR 0x0010
1595 #define MOPT_GTE0 0x0020
1596 #ifdef CONFIG_QUOTA
1597 #define MOPT_Q 0
1598 #define MOPT_QFMT 0x0040
1599 #else
1600 #define MOPT_Q MOPT_NOSUPPORT
1601 #define MOPT_QFMT MOPT_NOSUPPORT
1602 #endif
1603 #define MOPT_DATAJ 0x0080
1604 #define MOPT_NO_EXT2 0x0100
1605 #define MOPT_NO_EXT3 0x0200
1606 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1607 #define MOPT_STRING 0x0400
1608
1609 static const struct mount_opts {
1610 int token;
1611 int mount_opt;
1612 int flags;
1613 } ext4_mount_opts[] = {
1614 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1615 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1616 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1617 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1618 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1619 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1620 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1621 MOPT_EXT4_ONLY | MOPT_SET},
1622 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1623 MOPT_EXT4_ONLY | MOPT_CLEAR},
1624 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1625 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1626 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1627 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1628 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1629 MOPT_EXT4_ONLY | MOPT_CLEAR},
1630 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1631 MOPT_EXT4_ONLY | MOPT_CLEAR},
1632 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1633 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1634 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1635 EXT4_MOUNT_JOURNAL_CHECKSUM),
1636 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1637 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1638 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1639 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1640 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1641 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1642 MOPT_NO_EXT2},
1643 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1644 MOPT_NO_EXT2},
1645 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1646 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1647 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1648 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1649 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1650 {Opt_commit, 0, MOPT_GTE0},
1651 {Opt_max_batch_time, 0, MOPT_GTE0},
1652 {Opt_min_batch_time, 0, MOPT_GTE0},
1653 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1654 {Opt_init_itable, 0, MOPT_GTE0},
1655 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1656 {Opt_stripe, 0, MOPT_GTE0},
1657 {Opt_resuid, 0, MOPT_GTE0},
1658 {Opt_resgid, 0, MOPT_GTE0},
1659 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1660 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1661 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1662 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1663 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1664 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1665 MOPT_NO_EXT2 | MOPT_DATAJ},
1666 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1667 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1668 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1669 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1670 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1671 #else
1672 {Opt_acl, 0, MOPT_NOSUPPORT},
1673 {Opt_noacl, 0, MOPT_NOSUPPORT},
1674 #endif
1675 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1676 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1677 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1678 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1679 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1680 MOPT_SET | MOPT_Q},
1681 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1682 MOPT_SET | MOPT_Q},
1683 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1684 MOPT_SET | MOPT_Q},
1685 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1686 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1687 MOPT_CLEAR | MOPT_Q},
1688 {Opt_usrjquota, 0, MOPT_Q},
1689 {Opt_grpjquota, 0, MOPT_Q},
1690 {Opt_offusrjquota, 0, MOPT_Q},
1691 {Opt_offgrpjquota, 0, MOPT_Q},
1692 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1693 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1694 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1695 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1696 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1697 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1698 {Opt_err, 0, 0}
1699 };
1700
1701 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1702 substring_t *args, unsigned long *journal_devnum,
1703 unsigned int *journal_ioprio, int is_remount)
1704 {
1705 struct ext4_sb_info *sbi = EXT4_SB(sb);
1706 const struct mount_opts *m;
1707 kuid_t uid;
1708 kgid_t gid;
1709 int arg = 0;
1710
1711 #ifdef CONFIG_QUOTA
1712 if (token == Opt_usrjquota)
1713 return set_qf_name(sb, USRQUOTA, &args[0]);
1714 else if (token == Opt_grpjquota)
1715 return set_qf_name(sb, GRPQUOTA, &args[0]);
1716 else if (token == Opt_offusrjquota)
1717 return clear_qf_name(sb, USRQUOTA);
1718 else if (token == Opt_offgrpjquota)
1719 return clear_qf_name(sb, GRPQUOTA);
1720 #endif
1721 switch (token) {
1722 case Opt_noacl:
1723 case Opt_nouser_xattr:
1724 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1725 break;
1726 case Opt_sb:
1727 return 1; /* handled by get_sb_block() */
1728 case Opt_removed:
1729 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1730 return 1;
1731 case Opt_abort:
1732 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1733 return 1;
1734 case Opt_i_version:
1735 sb->s_flags |= SB_I_VERSION;
1736 return 1;
1737 case Opt_lazytime:
1738 sb->s_flags |= SB_LAZYTIME;
1739 return 1;
1740 case Opt_nolazytime:
1741 sb->s_flags &= ~SB_LAZYTIME;
1742 return 1;
1743 }
1744
1745 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1746 if (token == m->token)
1747 break;
1748
1749 if (m->token == Opt_err) {
1750 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1751 "or missing value", opt);
1752 return -1;
1753 }
1754
1755 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1756 ext4_msg(sb, KERN_ERR,
1757 "Mount option \"%s\" incompatible with ext2", opt);
1758 return -1;
1759 }
1760 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1761 ext4_msg(sb, KERN_ERR,
1762 "Mount option \"%s\" incompatible with ext3", opt);
1763 return -1;
1764 }
1765
1766 if (token == Opt_err_panic && !capable(CAP_SYS_ADMIN)) {
1767 ext4_msg(sb, KERN_ERR,
1768 "Mount option \"%s\" not allowed for unprivileged mounts",
1769 opt);
1770 return -1;
1771 }
1772
1773 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1774 return -1;
1775 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1776 return -1;
1777 if (m->flags & MOPT_EXPLICIT) {
1778 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1779 set_opt2(sb, EXPLICIT_DELALLOC);
1780 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1781 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1782 } else
1783 return -1;
1784 }
1785 if (m->flags & MOPT_CLEAR_ERR)
1786 clear_opt(sb, ERRORS_MASK);
1787 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1788 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1789 "options when quota turned on");
1790 return -1;
1791 }
1792
1793 if (m->flags & MOPT_NOSUPPORT) {
1794 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1795 } else if (token == Opt_commit) {
1796 if (arg == 0)
1797 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1798 sbi->s_commit_interval = HZ * arg;
1799 } else if (token == Opt_debug_want_extra_isize) {
1800 sbi->s_want_extra_isize = arg;
1801 } else if (token == Opt_max_batch_time) {
1802 sbi->s_max_batch_time = arg;
1803 } else if (token == Opt_min_batch_time) {
1804 sbi->s_min_batch_time = arg;
1805 } else if (token == Opt_inode_readahead_blks) {
1806 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1807 ext4_msg(sb, KERN_ERR,
1808 "EXT4-fs: inode_readahead_blks must be "
1809 "0 or a power of 2 smaller than 2^31");
1810 return -1;
1811 }
1812 sbi->s_inode_readahead_blks = arg;
1813 } else if (token == Opt_init_itable) {
1814 set_opt(sb, INIT_INODE_TABLE);
1815 if (!args->from)
1816 arg = EXT4_DEF_LI_WAIT_MULT;
1817 sbi->s_li_wait_mult = arg;
1818 } else if (token == Opt_max_dir_size_kb) {
1819 sbi->s_max_dir_size_kb = arg;
1820 } else if (token == Opt_stripe) {
1821 sbi->s_stripe = arg;
1822 } else if (token == Opt_resuid) {
1823 uid = make_kuid(sb->s_user_ns, arg);
1824 if (!uid_valid(uid)) {
1825 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1826 return -1;
1827 }
1828 sbi->s_resuid = uid;
1829 } else if (token == Opt_resgid) {
1830 gid = make_kgid(sb->s_user_ns, arg);
1831 if (!gid_valid(gid)) {
1832 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1833 return -1;
1834 }
1835 sbi->s_resgid = gid;
1836 } else if (token == Opt_journal_dev) {
1837 if (is_remount) {
1838 ext4_msg(sb, KERN_ERR,
1839 "Cannot specify journal on remount");
1840 return -1;
1841 }
1842 *journal_devnum = arg;
1843 } else if (token == Opt_journal_path) {
1844 char *journal_path;
1845 struct inode *journal_inode;
1846 struct path path;
1847 int error;
1848
1849 if (is_remount) {
1850 ext4_msg(sb, KERN_ERR,
1851 "Cannot specify journal on remount");
1852 return -1;
1853 }
1854 journal_path = match_strdup(&args[0]);
1855 if (!journal_path) {
1856 ext4_msg(sb, KERN_ERR, "error: could not dup "
1857 "journal device string");
1858 return -1;
1859 }
1860
1861 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1862 if (error) {
1863 ext4_msg(sb, KERN_ERR, "error: could not find "
1864 "journal device path: error %d", error);
1865 kfree(journal_path);
1866 return -1;
1867 }
1868
1869 /*
1870 * Refuse access for unprivileged mounts if the user does
1871 * not have rw access to the journal device via the supplied
1872 * path.
1873 */
1874 if (!capable(CAP_SYS_ADMIN) &&
1875 inode_permission(d_inode(path.dentry), MAY_READ|MAY_WRITE)) {
1876 ext4_msg(sb, KERN_ERR,
1877 "error: Insufficient access to journal path %s",
1878 journal_path);
1879 return -1;
1880 }
1881
1882 journal_inode = d_inode(path.dentry);
1883 if (!S_ISBLK(journal_inode->i_mode)) {
1884 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1885 "is not a block device", journal_path);
1886 path_put(&path);
1887 kfree(journal_path);
1888 return -1;
1889 }
1890
1891 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1892 path_put(&path);
1893 kfree(journal_path);
1894 } else if (token == Opt_journal_ioprio) {
1895 if (arg > 7) {
1896 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1897 " (must be 0-7)");
1898 return -1;
1899 }
1900 *journal_ioprio =
1901 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1902 } else if (token == Opt_test_dummy_encryption) {
1903 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1904 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1905 ext4_msg(sb, KERN_WARNING,
1906 "Test dummy encryption mode enabled");
1907 #else
1908 ext4_msg(sb, KERN_WARNING,
1909 "Test dummy encryption mount option ignored");
1910 #endif
1911 } else if (m->flags & MOPT_DATAJ) {
1912 if (is_remount) {
1913 if (!sbi->s_journal)
1914 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1915 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1916 ext4_msg(sb, KERN_ERR,
1917 "Cannot change data mode on remount");
1918 return -1;
1919 }
1920 } else {
1921 clear_opt(sb, DATA_FLAGS);
1922 sbi->s_mount_opt |= m->mount_opt;
1923 }
1924 #ifdef CONFIG_QUOTA
1925 } else if (m->flags & MOPT_QFMT) {
1926 if (sb_any_quota_loaded(sb) &&
1927 sbi->s_jquota_fmt != m->mount_opt) {
1928 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1929 "quota options when quota turned on");
1930 return -1;
1931 }
1932 if (ext4_has_feature_quota(sb)) {
1933 ext4_msg(sb, KERN_INFO,
1934 "Quota format mount options ignored "
1935 "when QUOTA feature is enabled");
1936 return 1;
1937 }
1938 sbi->s_jquota_fmt = m->mount_opt;
1939 #endif
1940 } else if (token == Opt_dax) {
1941 #ifdef CONFIG_FS_DAX
1942 ext4_msg(sb, KERN_WARNING,
1943 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1944 sbi->s_mount_opt |= m->mount_opt;
1945 #else
1946 ext4_msg(sb, KERN_INFO, "dax option not supported");
1947 return -1;
1948 #endif
1949 } else if (token == Opt_data_err_abort) {
1950 sbi->s_mount_opt |= m->mount_opt;
1951 } else if (token == Opt_data_err_ignore) {
1952 sbi->s_mount_opt &= ~m->mount_opt;
1953 } else {
1954 if (!args->from)
1955 arg = 1;
1956 if (m->flags & MOPT_CLEAR)
1957 arg = !arg;
1958 else if (unlikely(!(m->flags & MOPT_SET))) {
1959 ext4_msg(sb, KERN_WARNING,
1960 "buggy handling of option %s", opt);
1961 WARN_ON(1);
1962 return -1;
1963 }
1964 if (arg != 0)
1965 sbi->s_mount_opt |= m->mount_opt;
1966 else
1967 sbi->s_mount_opt &= ~m->mount_opt;
1968 }
1969 return 1;
1970 }
1971
1972 static int parse_options(char *options, struct super_block *sb,
1973 unsigned long *journal_devnum,
1974 unsigned int *journal_ioprio,
1975 int is_remount)
1976 {
1977 struct ext4_sb_info *sbi = EXT4_SB(sb);
1978 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
1979 substring_t args[MAX_OPT_ARGS];
1980 int token;
1981
1982 if (!options)
1983 return 1;
1984
1985 while ((p = strsep(&options, ",")) != NULL) {
1986 if (!*p)
1987 continue;
1988 /*
1989 * Initialize args struct so we know whether arg was
1990 * found; some options take optional arguments.
1991 */
1992 args[0].to = args[0].from = NULL;
1993 token = match_token(p, tokens, args);
1994 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1995 journal_ioprio, is_remount) < 0)
1996 return 0;
1997 }
1998 #ifdef CONFIG_QUOTA
1999 /*
2000 * We do the test below only for project quotas. 'usrquota' and
2001 * 'grpquota' mount options are allowed even without quota feature
2002 * to support legacy quotas in quota files.
2003 */
2004 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2005 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2006 "Cannot enable project quota enforcement.");
2007 return 0;
2008 }
2009 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2010 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2011 if (usr_qf_name || grp_qf_name) {
2012 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2013 clear_opt(sb, USRQUOTA);
2014
2015 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2016 clear_opt(sb, GRPQUOTA);
2017
2018 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2019 ext4_msg(sb, KERN_ERR, "old and new quota "
2020 "format mixing");
2021 return 0;
2022 }
2023
2024 if (!sbi->s_jquota_fmt) {
2025 ext4_msg(sb, KERN_ERR, "journaled quota format "
2026 "not specified");
2027 return 0;
2028 }
2029 }
2030 #endif
2031 if (test_opt(sb, DIOREAD_NOLOCK)) {
2032 int blocksize =
2033 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2034
2035 if (blocksize < PAGE_SIZE) {
2036 ext4_msg(sb, KERN_ERR, "can't mount with "
2037 "dioread_nolock if block size != PAGE_SIZE");
2038 return 0;
2039 }
2040 }
2041 return 1;
2042 }
2043
2044 static inline void ext4_show_quota_options(struct seq_file *seq,
2045 struct super_block *sb)
2046 {
2047 #if defined(CONFIG_QUOTA)
2048 struct ext4_sb_info *sbi = EXT4_SB(sb);
2049 char *usr_qf_name, *grp_qf_name;
2050
2051 if (sbi->s_jquota_fmt) {
2052 char *fmtname = "";
2053
2054 switch (sbi->s_jquota_fmt) {
2055 case QFMT_VFS_OLD:
2056 fmtname = "vfsold";
2057 break;
2058 case QFMT_VFS_V0:
2059 fmtname = "vfsv0";
2060 break;
2061 case QFMT_VFS_V1:
2062 fmtname = "vfsv1";
2063 break;
2064 }
2065 seq_printf(seq, ",jqfmt=%s", fmtname);
2066 }
2067
2068 rcu_read_lock();
2069 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2070 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2071 if (usr_qf_name)
2072 seq_show_option(seq, "usrjquota", usr_qf_name);
2073 if (grp_qf_name)
2074 seq_show_option(seq, "grpjquota", grp_qf_name);
2075 rcu_read_unlock();
2076 #endif
2077 }
2078
2079 static const char *token2str(int token)
2080 {
2081 const struct match_token *t;
2082
2083 for (t = tokens; t->token != Opt_err; t++)
2084 if (t->token == token && !strchr(t->pattern, '='))
2085 break;
2086 return t->pattern;
2087 }
2088
2089 /*
2090 * Show an option if
2091 * - it's set to a non-default value OR
2092 * - if the per-sb default is different from the global default
2093 */
2094 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2095 int nodefs)
2096 {
2097 struct ext4_sb_info *sbi = EXT4_SB(sb);
2098 struct ext4_super_block *es = sbi->s_es;
2099 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
2100 const struct mount_opts *m;
2101 char sep = nodefs ? '\n' : ',';
2102
2103 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2104 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2105
2106 if (sbi->s_sb_block != 1)
2107 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2108
2109 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2110 int want_set = m->flags & MOPT_SET;
2111 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2112 (m->flags & MOPT_CLEAR_ERR))
2113 continue;
2114 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2115 continue; /* skip if same as the default */
2116 if ((want_set &&
2117 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2118 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2119 continue; /* select Opt_noFoo vs Opt_Foo */
2120 SEQ_OPTS_PRINT("%s", token2str(m->token));
2121 }
2122
2123 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(sb->s_user_ns, EXT4_DEF_RESUID)) ||
2124 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2125 SEQ_OPTS_PRINT("resuid=%u",
2126 from_kuid_munged(sb->s_user_ns, sbi->s_resuid));
2127 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(sb->s_user_ns, EXT4_DEF_RESGID)) ||
2128 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2129 SEQ_OPTS_PRINT("resgid=%u",
2130 from_kgid_munged(sb->s_user_ns, sbi->s_resgid));
2131 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2132 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2133 SEQ_OPTS_PUTS("errors=remount-ro");
2134 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2135 SEQ_OPTS_PUTS("errors=continue");
2136 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2137 SEQ_OPTS_PUTS("errors=panic");
2138 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2139 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2140 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2141 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2142 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2143 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2144 if (sb->s_flags & SB_I_VERSION)
2145 SEQ_OPTS_PUTS("i_version");
2146 if (nodefs || sbi->s_stripe)
2147 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2148 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
2149 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2150 SEQ_OPTS_PUTS("data=journal");
2151 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2152 SEQ_OPTS_PUTS("data=ordered");
2153 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2154 SEQ_OPTS_PUTS("data=writeback");
2155 }
2156 if (nodefs ||
2157 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2158 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2159 sbi->s_inode_readahead_blks);
2160
2161 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2162 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2163 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2164 if (nodefs || sbi->s_max_dir_size_kb)
2165 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2166 if (test_opt(sb, DATA_ERR_ABORT))
2167 SEQ_OPTS_PUTS("data_err=abort");
2168 if (DUMMY_ENCRYPTION_ENABLED(sbi))
2169 SEQ_OPTS_PUTS("test_dummy_encryption");
2170
2171 ext4_show_quota_options(seq, sb);
2172 return 0;
2173 }
2174
2175 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2176 {
2177 return _ext4_show_options(seq, root->d_sb, 0);
2178 }
2179
2180 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2181 {
2182 struct super_block *sb = seq->private;
2183 int rc;
2184
2185 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2186 rc = _ext4_show_options(seq, sb, 1);
2187 seq_puts(seq, "\n");
2188 return rc;
2189 }
2190
2191 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2192 int read_only)
2193 {
2194 struct ext4_sb_info *sbi = EXT4_SB(sb);
2195 int res = 0;
2196
2197 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2198 ext4_msg(sb, KERN_ERR, "revision level too high, "
2199 "forcing read-only mode");
2200 res = SB_RDONLY;
2201 }
2202 if (read_only)
2203 goto done;
2204 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2205 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2206 "running e2fsck is recommended");
2207 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2208 ext4_msg(sb, KERN_WARNING,
2209 "warning: mounting fs with errors, "
2210 "running e2fsck is recommended");
2211 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2212 le16_to_cpu(es->s_mnt_count) >=
2213 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2214 ext4_msg(sb, KERN_WARNING,
2215 "warning: maximal mount count reached, "
2216 "running e2fsck is recommended");
2217 else if (le32_to_cpu(es->s_checkinterval) &&
2218 (le32_to_cpu(es->s_lastcheck) +
2219 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2220 ext4_msg(sb, KERN_WARNING,
2221 "warning: checktime reached, "
2222 "running e2fsck is recommended");
2223 if (!sbi->s_journal)
2224 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2225 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2226 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2227 le16_add_cpu(&es->s_mnt_count, 1);
2228 es->s_mtime = cpu_to_le32(get_seconds());
2229 ext4_update_dynamic_rev(sb);
2230 if (sbi->s_journal)
2231 ext4_set_feature_journal_needs_recovery(sb);
2232
2233 ext4_commit_super(sb, 1);
2234 done:
2235 if (test_opt(sb, DEBUG))
2236 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2237 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2238 sb->s_blocksize,
2239 sbi->s_groups_count,
2240 EXT4_BLOCKS_PER_GROUP(sb),
2241 EXT4_INODES_PER_GROUP(sb),
2242 sbi->s_mount_opt, sbi->s_mount_opt2);
2243
2244 cleancache_init_fs(sb);
2245 return res;
2246 }
2247
2248 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2249 {
2250 struct ext4_sb_info *sbi = EXT4_SB(sb);
2251 struct flex_groups *new_groups;
2252 int size;
2253
2254 if (!sbi->s_log_groups_per_flex)
2255 return 0;
2256
2257 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2258 if (size <= sbi->s_flex_groups_allocated)
2259 return 0;
2260
2261 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2262 new_groups = kvzalloc(size, GFP_KERNEL);
2263 if (!new_groups) {
2264 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2265 size / (int) sizeof(struct flex_groups));
2266 return -ENOMEM;
2267 }
2268
2269 if (sbi->s_flex_groups) {
2270 memcpy(new_groups, sbi->s_flex_groups,
2271 (sbi->s_flex_groups_allocated *
2272 sizeof(struct flex_groups)));
2273 kvfree(sbi->s_flex_groups);
2274 }
2275 sbi->s_flex_groups = new_groups;
2276 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2277 return 0;
2278 }
2279
2280 static int ext4_fill_flex_info(struct super_block *sb)
2281 {
2282 struct ext4_sb_info *sbi = EXT4_SB(sb);
2283 struct ext4_group_desc *gdp = NULL;
2284 ext4_group_t flex_group;
2285 int i, err;
2286
2287 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2288 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2289 sbi->s_log_groups_per_flex = 0;
2290 return 1;
2291 }
2292
2293 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2294 if (err)
2295 goto failed;
2296
2297 for (i = 0; i < sbi->s_groups_count; i++) {
2298 gdp = ext4_get_group_desc(sb, i, NULL);
2299
2300 flex_group = ext4_flex_group(sbi, i);
2301 atomic_add(ext4_free_inodes_count(sb, gdp),
2302 &sbi->s_flex_groups[flex_group].free_inodes);
2303 atomic64_add(ext4_free_group_clusters(sb, gdp),
2304 &sbi->s_flex_groups[flex_group].free_clusters);
2305 atomic_add(ext4_used_dirs_count(sb, gdp),
2306 &sbi->s_flex_groups[flex_group].used_dirs);
2307 }
2308
2309 return 1;
2310 failed:
2311 return 0;
2312 }
2313
2314 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2315 struct ext4_group_desc *gdp)
2316 {
2317 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2318 __u16 crc = 0;
2319 __le32 le_group = cpu_to_le32(block_group);
2320 struct ext4_sb_info *sbi = EXT4_SB(sb);
2321
2322 if (ext4_has_metadata_csum(sbi->s_sb)) {
2323 /* Use new metadata_csum algorithm */
2324 __u32 csum32;
2325 __u16 dummy_csum = 0;
2326
2327 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2328 sizeof(le_group));
2329 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2330 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2331 sizeof(dummy_csum));
2332 offset += sizeof(dummy_csum);
2333 if (offset < sbi->s_desc_size)
2334 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2335 sbi->s_desc_size - offset);
2336
2337 crc = csum32 & 0xFFFF;
2338 goto out;
2339 }
2340
2341 /* old crc16 code */
2342 if (!ext4_has_feature_gdt_csum(sb))
2343 return 0;
2344
2345 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2346 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2347 crc = crc16(crc, (__u8 *)gdp, offset);
2348 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2349 /* for checksum of struct ext4_group_desc do the rest...*/
2350 if (ext4_has_feature_64bit(sb) &&
2351 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2352 crc = crc16(crc, (__u8 *)gdp + offset,
2353 le16_to_cpu(sbi->s_es->s_desc_size) -
2354 offset);
2355
2356 out:
2357 return cpu_to_le16(crc);
2358 }
2359
2360 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2361 struct ext4_group_desc *gdp)
2362 {
2363 if (ext4_has_group_desc_csum(sb) &&
2364 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2365 return 0;
2366
2367 return 1;
2368 }
2369
2370 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2371 struct ext4_group_desc *gdp)
2372 {
2373 if (!ext4_has_group_desc_csum(sb))
2374 return;
2375 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2376 }
2377
2378 /* Called at mount-time, super-block is locked */
2379 static int ext4_check_descriptors(struct super_block *sb,
2380 ext4_fsblk_t sb_block,
2381 ext4_group_t *first_not_zeroed)
2382 {
2383 struct ext4_sb_info *sbi = EXT4_SB(sb);
2384 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2385 ext4_fsblk_t last_block;
2386 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2387 ext4_fsblk_t block_bitmap;
2388 ext4_fsblk_t inode_bitmap;
2389 ext4_fsblk_t inode_table;
2390 int flexbg_flag = 0;
2391 ext4_group_t i, grp = sbi->s_groups_count;
2392
2393 if (ext4_has_feature_flex_bg(sb))
2394 flexbg_flag = 1;
2395
2396 ext4_debug("Checking group descriptors");
2397
2398 for (i = 0; i < sbi->s_groups_count; i++) {
2399 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2400
2401 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2402 last_block = ext4_blocks_count(sbi->s_es) - 1;
2403 else
2404 last_block = first_block +
2405 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2406
2407 if ((grp == sbi->s_groups_count) &&
2408 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2409 grp = i;
2410
2411 block_bitmap = ext4_block_bitmap(sb, gdp);
2412 if (block_bitmap == sb_block) {
2413 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2414 "Block bitmap for group %u overlaps "
2415 "superblock", i);
2416 if (!sb_rdonly(sb))
2417 return 0;
2418 }
2419 if (block_bitmap >= sb_block + 1 &&
2420 block_bitmap <= last_bg_block) {
2421 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2422 "Block bitmap for group %u overlaps "
2423 "block group descriptors", i);
2424 if (!sb_rdonly(sb))
2425 return 0;
2426 }
2427 if (block_bitmap < first_block || block_bitmap > last_block) {
2428 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2429 "Block bitmap for group %u not in group "
2430 "(block %llu)!", i, block_bitmap);
2431 return 0;
2432 }
2433 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2434 if (inode_bitmap == sb_block) {
2435 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2436 "Inode bitmap for group %u overlaps "
2437 "superblock", i);
2438 if (!sb_rdonly(sb))
2439 return 0;
2440 }
2441 if (inode_bitmap >= sb_block + 1 &&
2442 inode_bitmap <= last_bg_block) {
2443 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2444 "Inode bitmap for group %u overlaps "
2445 "block group descriptors", i);
2446 if (!sb_rdonly(sb))
2447 return 0;
2448 }
2449 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2450 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2451 "Inode bitmap for group %u not in group "
2452 "(block %llu)!", i, inode_bitmap);
2453 return 0;
2454 }
2455 inode_table = ext4_inode_table(sb, gdp);
2456 if (inode_table == sb_block) {
2457 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2458 "Inode table for group %u overlaps "
2459 "superblock", i);
2460 if (!sb_rdonly(sb))
2461 return 0;
2462 }
2463 if (inode_table >= sb_block + 1 &&
2464 inode_table <= last_bg_block) {
2465 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2466 "Inode table for group %u overlaps "
2467 "block group descriptors", i);
2468 if (!sb_rdonly(sb))
2469 return 0;
2470 }
2471 if (inode_table < first_block ||
2472 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2473 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2474 "Inode table for group %u not in group "
2475 "(block %llu)!", i, inode_table);
2476 return 0;
2477 }
2478 ext4_lock_group(sb, i);
2479 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2480 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2481 "Checksum for group %u failed (%u!=%u)",
2482 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2483 gdp)), le16_to_cpu(gdp->bg_checksum));
2484 if (!sb_rdonly(sb)) {
2485 ext4_unlock_group(sb, i);
2486 return 0;
2487 }
2488 }
2489 ext4_unlock_group(sb, i);
2490 if (!flexbg_flag)
2491 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2492 }
2493 if (NULL != first_not_zeroed)
2494 *first_not_zeroed = grp;
2495 return 1;
2496 }
2497
2498 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2499 * the superblock) which were deleted from all directories, but held open by
2500 * a process at the time of a crash. We walk the list and try to delete these
2501 * inodes at recovery time (only with a read-write filesystem).
2502 *
2503 * In order to keep the orphan inode chain consistent during traversal (in
2504 * case of crash during recovery), we link each inode into the superblock
2505 * orphan list_head and handle it the same way as an inode deletion during
2506 * normal operation (which journals the operations for us).
2507 *
2508 * We only do an iget() and an iput() on each inode, which is very safe if we
2509 * accidentally point at an in-use or already deleted inode. The worst that
2510 * can happen in this case is that we get a "bit already cleared" message from
2511 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2512 * e2fsck was run on this filesystem, and it must have already done the orphan
2513 * inode cleanup for us, so we can safely abort without any further action.
2514 */
2515 static void ext4_orphan_cleanup(struct super_block *sb,
2516 struct ext4_super_block *es)
2517 {
2518 unsigned int s_flags = sb->s_flags;
2519 int ret, nr_orphans = 0, nr_truncates = 0;
2520 #ifdef CONFIG_QUOTA
2521 int quota_update = 0;
2522 int i;
2523 #endif
2524 if (!es->s_last_orphan) {
2525 jbd_debug(4, "no orphan inodes to clean up\n");
2526 return;
2527 }
2528
2529 if (bdev_read_only(sb->s_bdev)) {
2530 ext4_msg(sb, KERN_ERR, "write access "
2531 "unavailable, skipping orphan cleanup");
2532 return;
2533 }
2534
2535 /* Check if feature set would not allow a r/w mount */
2536 if (!ext4_feature_set_ok(sb, 0)) {
2537 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2538 "unknown ROCOMPAT features");
2539 return;
2540 }
2541
2542 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2543 /* don't clear list on RO mount w/ errors */
2544 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2545 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2546 "clearing orphan list.\n");
2547 es->s_last_orphan = 0;
2548 }
2549 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2550 return;
2551 }
2552
2553 if (s_flags & SB_RDONLY) {
2554 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2555 sb->s_flags &= ~SB_RDONLY;
2556 }
2557 #ifdef CONFIG_QUOTA
2558 /* Needed for iput() to work correctly and not trash data */
2559 sb->s_flags |= SB_ACTIVE;
2560
2561 /*
2562 * Turn on quotas which were not enabled for read-only mounts if
2563 * filesystem has quota feature, so that they are updated correctly.
2564 */
2565 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
2566 int ret = ext4_enable_quotas(sb);
2567
2568 if (!ret)
2569 quota_update = 1;
2570 else
2571 ext4_msg(sb, KERN_ERR,
2572 "Cannot turn on quotas: error %d", ret);
2573 }
2574
2575 /* Turn on journaled quotas used for old sytle */
2576 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2577 if (EXT4_SB(sb)->s_qf_names[i]) {
2578 int ret = ext4_quota_on_mount(sb, i);
2579
2580 if (!ret)
2581 quota_update = 1;
2582 else
2583 ext4_msg(sb, KERN_ERR,
2584 "Cannot turn on journaled "
2585 "quota: type %d: error %d", i, ret);
2586 }
2587 }
2588 #endif
2589
2590 while (es->s_last_orphan) {
2591 struct inode *inode;
2592
2593 /*
2594 * We may have encountered an error during cleanup; if
2595 * so, skip the rest.
2596 */
2597 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2598 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2599 es->s_last_orphan = 0;
2600 break;
2601 }
2602
2603 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2604 if (IS_ERR(inode)) {
2605 es->s_last_orphan = 0;
2606 break;
2607 }
2608
2609 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2610 dquot_initialize(inode);
2611 if (inode->i_nlink) {
2612 if (test_opt(sb, DEBUG))
2613 ext4_msg(sb, KERN_DEBUG,
2614 "%s: truncating inode %lu to %lld bytes",
2615 __func__, inode->i_ino, inode->i_size);
2616 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2617 inode->i_ino, inode->i_size);
2618 inode_lock(inode);
2619 truncate_inode_pages(inode->i_mapping, inode->i_size);
2620 ret = ext4_truncate(inode);
2621 if (ret)
2622 ext4_std_error(inode->i_sb, ret);
2623 inode_unlock(inode);
2624 nr_truncates++;
2625 } else {
2626 if (test_opt(sb, DEBUG))
2627 ext4_msg(sb, KERN_DEBUG,
2628 "%s: deleting unreferenced inode %lu",
2629 __func__, inode->i_ino);
2630 jbd_debug(2, "deleting unreferenced inode %lu\n",
2631 inode->i_ino);
2632 nr_orphans++;
2633 }
2634 iput(inode); /* The delete magic happens here! */
2635 }
2636
2637 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2638
2639 if (nr_orphans)
2640 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2641 PLURAL(nr_orphans));
2642 if (nr_truncates)
2643 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2644 PLURAL(nr_truncates));
2645 #ifdef CONFIG_QUOTA
2646 /* Turn off quotas if they were enabled for orphan cleanup */
2647 if (quota_update) {
2648 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2649 if (sb_dqopt(sb)->files[i])
2650 dquot_quota_off(sb, i);
2651 }
2652 }
2653 #endif
2654 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2655 }
2656
2657 /*
2658 * Maximal extent format file size.
2659 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2660 * extent format containers, within a sector_t, and within i_blocks
2661 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2662 * so that won't be a limiting factor.
2663 *
2664 * However there is other limiting factor. We do store extents in the form
2665 * of starting block and length, hence the resulting length of the extent
2666 * covering maximum file size must fit into on-disk format containers as
2667 * well. Given that length is always by 1 unit bigger than max unit (because
2668 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2669 *
2670 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2671 */
2672 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2673 {
2674 loff_t res;
2675 loff_t upper_limit = MAX_LFS_FILESIZE;
2676
2677 /* small i_blocks in vfs inode? */
2678 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2679 /*
2680 * CONFIG_LBDAF is not enabled implies the inode
2681 * i_block represent total blocks in 512 bytes
2682 * 32 == size of vfs inode i_blocks * 8
2683 */
2684 upper_limit = (1LL << 32) - 1;
2685
2686 /* total blocks in file system block size */
2687 upper_limit >>= (blkbits - 9);
2688 upper_limit <<= blkbits;
2689 }
2690
2691 /*
2692 * 32-bit extent-start container, ee_block. We lower the maxbytes
2693 * by one fs block, so ee_len can cover the extent of maximum file
2694 * size
2695 */
2696 res = (1LL << 32) - 1;
2697 res <<= blkbits;
2698
2699 /* Sanity check against vm- & vfs- imposed limits */
2700 if (res > upper_limit)
2701 res = upper_limit;
2702
2703 return res;
2704 }
2705
2706 /*
2707 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2708 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2709 * We need to be 1 filesystem block less than the 2^48 sector limit.
2710 */
2711 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2712 {
2713 loff_t res = EXT4_NDIR_BLOCKS;
2714 int meta_blocks;
2715 loff_t upper_limit;
2716 /* This is calculated to be the largest file size for a dense, block
2717 * mapped file such that the file's total number of 512-byte sectors,
2718 * including data and all indirect blocks, does not exceed (2^48 - 1).
2719 *
2720 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2721 * number of 512-byte sectors of the file.
2722 */
2723
2724 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2725 /*
2726 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2727 * the inode i_block field represents total file blocks in
2728 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2729 */
2730 upper_limit = (1LL << 32) - 1;
2731
2732 /* total blocks in file system block size */
2733 upper_limit >>= (bits - 9);
2734
2735 } else {
2736 /*
2737 * We use 48 bit ext4_inode i_blocks
2738 * With EXT4_HUGE_FILE_FL set the i_blocks
2739 * represent total number of blocks in
2740 * file system block size
2741 */
2742 upper_limit = (1LL << 48) - 1;
2743
2744 }
2745
2746 /* indirect blocks */
2747 meta_blocks = 1;
2748 /* double indirect blocks */
2749 meta_blocks += 1 + (1LL << (bits-2));
2750 /* tripple indirect blocks */
2751 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2752
2753 upper_limit -= meta_blocks;
2754 upper_limit <<= bits;
2755
2756 res += 1LL << (bits-2);
2757 res += 1LL << (2*(bits-2));
2758 res += 1LL << (3*(bits-2));
2759 res <<= bits;
2760 if (res > upper_limit)
2761 res = upper_limit;
2762
2763 if (res > MAX_LFS_FILESIZE)
2764 res = MAX_LFS_FILESIZE;
2765
2766 return res;
2767 }
2768
2769 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2770 ext4_fsblk_t logical_sb_block, int nr)
2771 {
2772 struct ext4_sb_info *sbi = EXT4_SB(sb);
2773 ext4_group_t bg, first_meta_bg;
2774 int has_super = 0;
2775
2776 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2777
2778 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2779 return logical_sb_block + nr + 1;
2780 bg = sbi->s_desc_per_block * nr;
2781 if (ext4_bg_has_super(sb, bg))
2782 has_super = 1;
2783
2784 /*
2785 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2786 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2787 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2788 * compensate.
2789 */
2790 if (sb->s_blocksize == 1024 && nr == 0 &&
2791 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2792 has_super++;
2793
2794 return (has_super + ext4_group_first_block_no(sb, bg));
2795 }
2796
2797 /**
2798 * ext4_get_stripe_size: Get the stripe size.
2799 * @sbi: In memory super block info
2800 *
2801 * If we have specified it via mount option, then
2802 * use the mount option value. If the value specified at mount time is
2803 * greater than the blocks per group use the super block value.
2804 * If the super block value is greater than blocks per group return 0.
2805 * Allocator needs it be less than blocks per group.
2806 *
2807 */
2808 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2809 {
2810 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2811 unsigned long stripe_width =
2812 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2813 int ret;
2814
2815 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2816 ret = sbi->s_stripe;
2817 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2818 ret = stripe_width;
2819 else if (stride && stride <= sbi->s_blocks_per_group)
2820 ret = stride;
2821 else
2822 ret = 0;
2823
2824 /*
2825 * If the stripe width is 1, this makes no sense and
2826 * we set it to 0 to turn off stripe handling code.
2827 */
2828 if (ret <= 1)
2829 ret = 0;
2830
2831 return ret;
2832 }
2833
2834 /*
2835 * Check whether this filesystem can be mounted based on
2836 * the features present and the RDONLY/RDWR mount requested.
2837 * Returns 1 if this filesystem can be mounted as requested,
2838 * 0 if it cannot be.
2839 */
2840 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2841 {
2842 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2843 ext4_msg(sb, KERN_ERR,
2844 "Couldn't mount because of "
2845 "unsupported optional features (%x)",
2846 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2847 ~EXT4_FEATURE_INCOMPAT_SUPP));
2848 return 0;
2849 }
2850
2851 if (readonly)
2852 return 1;
2853
2854 if (ext4_has_feature_readonly(sb)) {
2855 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2856 sb->s_flags |= SB_RDONLY;
2857 return 1;
2858 }
2859
2860 /* Check that feature set is OK for a read-write mount */
2861 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2862 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2863 "unsupported optional features (%x)",
2864 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2865 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2866 return 0;
2867 }
2868 /*
2869 * Large file size enabled file system can only be mounted
2870 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2871 */
2872 if (ext4_has_feature_huge_file(sb)) {
2873 if (sizeof(blkcnt_t) < sizeof(u64)) {
2874 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2875 "cannot be mounted RDWR without "
2876 "CONFIG_LBDAF");
2877 return 0;
2878 }
2879 }
2880 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2881 ext4_msg(sb, KERN_ERR,
2882 "Can't support bigalloc feature without "
2883 "extents feature\n");
2884 return 0;
2885 }
2886
2887 #ifndef CONFIG_QUOTA
2888 if (ext4_has_feature_quota(sb) && !readonly) {
2889 ext4_msg(sb, KERN_ERR,
2890 "Filesystem with quota feature cannot be mounted RDWR "
2891 "without CONFIG_QUOTA");
2892 return 0;
2893 }
2894 if (ext4_has_feature_project(sb) && !readonly) {
2895 ext4_msg(sb, KERN_ERR,
2896 "Filesystem with project quota feature cannot be mounted RDWR "
2897 "without CONFIG_QUOTA");
2898 return 0;
2899 }
2900 #endif /* CONFIG_QUOTA */
2901 return 1;
2902 }
2903
2904 /*
2905 * This function is called once a day if we have errors logged
2906 * on the file system
2907 */
2908 static void print_daily_error_info(struct timer_list *t)
2909 {
2910 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
2911 struct super_block *sb = sbi->s_sb;
2912 struct ext4_super_block *es = sbi->s_es;
2913
2914 if (es->s_error_count)
2915 /* fsck newer than v1.41.13 is needed to clean this condition. */
2916 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2917 le32_to_cpu(es->s_error_count));
2918 if (es->s_first_error_time) {
2919 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2920 sb->s_id, le32_to_cpu(es->s_first_error_time),
2921 (int) sizeof(es->s_first_error_func),
2922 es->s_first_error_func,
2923 le32_to_cpu(es->s_first_error_line));
2924 if (es->s_first_error_ino)
2925 printk(KERN_CONT ": inode %u",
2926 le32_to_cpu(es->s_first_error_ino));
2927 if (es->s_first_error_block)
2928 printk(KERN_CONT ": block %llu", (unsigned long long)
2929 le64_to_cpu(es->s_first_error_block));
2930 printk(KERN_CONT "\n");
2931 }
2932 if (es->s_last_error_time) {
2933 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2934 sb->s_id, le32_to_cpu(es->s_last_error_time),
2935 (int) sizeof(es->s_last_error_func),
2936 es->s_last_error_func,
2937 le32_to_cpu(es->s_last_error_line));
2938 if (es->s_last_error_ino)
2939 printk(KERN_CONT ": inode %u",
2940 le32_to_cpu(es->s_last_error_ino));
2941 if (es->s_last_error_block)
2942 printk(KERN_CONT ": block %llu", (unsigned long long)
2943 le64_to_cpu(es->s_last_error_block));
2944 printk(KERN_CONT "\n");
2945 }
2946 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2947 }
2948
2949 /* Find next suitable group and run ext4_init_inode_table */
2950 static int ext4_run_li_request(struct ext4_li_request *elr)
2951 {
2952 struct ext4_group_desc *gdp = NULL;
2953 ext4_group_t group, ngroups;
2954 struct super_block *sb;
2955 unsigned long timeout = 0;
2956 int ret = 0;
2957
2958 sb = elr->lr_super;
2959 ngroups = EXT4_SB(sb)->s_groups_count;
2960
2961 for (group = elr->lr_next_group; group < ngroups; group++) {
2962 gdp = ext4_get_group_desc(sb, group, NULL);
2963 if (!gdp) {
2964 ret = 1;
2965 break;
2966 }
2967
2968 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2969 break;
2970 }
2971
2972 if (group >= ngroups)
2973 ret = 1;
2974
2975 if (!ret) {
2976 timeout = jiffies;
2977 ret = ext4_init_inode_table(sb, group,
2978 elr->lr_timeout ? 0 : 1);
2979 if (elr->lr_timeout == 0) {
2980 timeout = (jiffies - timeout) *
2981 elr->lr_sbi->s_li_wait_mult;
2982 elr->lr_timeout = timeout;
2983 }
2984 elr->lr_next_sched = jiffies + elr->lr_timeout;
2985 elr->lr_next_group = group + 1;
2986 }
2987 return ret;
2988 }
2989
2990 /*
2991 * Remove lr_request from the list_request and free the
2992 * request structure. Should be called with li_list_mtx held
2993 */
2994 static void ext4_remove_li_request(struct ext4_li_request *elr)
2995 {
2996 struct ext4_sb_info *sbi;
2997
2998 if (!elr)
2999 return;
3000
3001 sbi = elr->lr_sbi;
3002
3003 list_del(&elr->lr_request);
3004 sbi->s_li_request = NULL;
3005 kfree(elr);
3006 }
3007
3008 static void ext4_unregister_li_request(struct super_block *sb)
3009 {
3010 mutex_lock(&ext4_li_mtx);
3011 if (!ext4_li_info) {
3012 mutex_unlock(&ext4_li_mtx);
3013 return;
3014 }
3015
3016 mutex_lock(&ext4_li_info->li_list_mtx);
3017 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3018 mutex_unlock(&ext4_li_info->li_list_mtx);
3019 mutex_unlock(&ext4_li_mtx);
3020 }
3021
3022 static struct task_struct *ext4_lazyinit_task;
3023
3024 /*
3025 * This is the function where ext4lazyinit thread lives. It walks
3026 * through the request list searching for next scheduled filesystem.
3027 * When such a fs is found, run the lazy initialization request
3028 * (ext4_rn_li_request) and keep track of the time spend in this
3029 * function. Based on that time we compute next schedule time of
3030 * the request. When walking through the list is complete, compute
3031 * next waking time and put itself into sleep.
3032 */
3033 static int ext4_lazyinit_thread(void *arg)
3034 {
3035 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3036 struct list_head *pos, *n;
3037 struct ext4_li_request *elr;
3038 unsigned long next_wakeup, cur;
3039
3040 BUG_ON(NULL == eli);
3041
3042 cont_thread:
3043 while (true) {
3044 next_wakeup = MAX_JIFFY_OFFSET;
3045
3046 mutex_lock(&eli->li_list_mtx);
3047 if (list_empty(&eli->li_request_list)) {
3048 mutex_unlock(&eli->li_list_mtx);
3049 goto exit_thread;
3050 }
3051 list_for_each_safe(pos, n, &eli->li_request_list) {
3052 int err = 0;
3053 int progress = 0;
3054 elr = list_entry(pos, struct ext4_li_request,
3055 lr_request);
3056
3057 if (time_before(jiffies, elr->lr_next_sched)) {
3058 if (time_before(elr->lr_next_sched, next_wakeup))
3059 next_wakeup = elr->lr_next_sched;
3060 continue;
3061 }
3062 if (down_read_trylock(&elr->lr_super->s_umount)) {
3063 if (sb_start_write_trylock(elr->lr_super)) {
3064 progress = 1;
3065 /*
3066 * We hold sb->s_umount, sb can not
3067 * be removed from the list, it is
3068 * now safe to drop li_list_mtx
3069 */
3070 mutex_unlock(&eli->li_list_mtx);
3071 err = ext4_run_li_request(elr);
3072 sb_end_write(elr->lr_super);
3073 mutex_lock(&eli->li_list_mtx);
3074 n = pos->next;
3075 }
3076 up_read((&elr->lr_super->s_umount));
3077 }
3078 /* error, remove the lazy_init job */
3079 if (err) {
3080 ext4_remove_li_request(elr);
3081 continue;
3082 }
3083 if (!progress) {
3084 elr->lr_next_sched = jiffies +
3085 (prandom_u32()
3086 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3087 }
3088 if (time_before(elr->lr_next_sched, next_wakeup))
3089 next_wakeup = elr->lr_next_sched;
3090 }
3091 mutex_unlock(&eli->li_list_mtx);
3092
3093 try_to_freeze();
3094
3095 cur = jiffies;
3096 if ((time_after_eq(cur, next_wakeup)) ||
3097 (MAX_JIFFY_OFFSET == next_wakeup)) {
3098 cond_resched();
3099 continue;
3100 }
3101
3102 schedule_timeout_interruptible(next_wakeup - cur);
3103
3104 if (kthread_should_stop()) {
3105 ext4_clear_request_list();
3106 goto exit_thread;
3107 }
3108 }
3109
3110 exit_thread:
3111 /*
3112 * It looks like the request list is empty, but we need
3113 * to check it under the li_list_mtx lock, to prevent any
3114 * additions into it, and of course we should lock ext4_li_mtx
3115 * to atomically free the list and ext4_li_info, because at
3116 * this point another ext4 filesystem could be registering
3117 * new one.
3118 */
3119 mutex_lock(&ext4_li_mtx);
3120 mutex_lock(&eli->li_list_mtx);
3121 if (!list_empty(&eli->li_request_list)) {
3122 mutex_unlock(&eli->li_list_mtx);
3123 mutex_unlock(&ext4_li_mtx);
3124 goto cont_thread;
3125 }
3126 mutex_unlock(&eli->li_list_mtx);
3127 kfree(ext4_li_info);
3128 ext4_li_info = NULL;
3129 mutex_unlock(&ext4_li_mtx);
3130
3131 return 0;
3132 }
3133
3134 static void ext4_clear_request_list(void)
3135 {
3136 struct list_head *pos, *n;
3137 struct ext4_li_request *elr;
3138
3139 mutex_lock(&ext4_li_info->li_list_mtx);
3140 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3141 elr = list_entry(pos, struct ext4_li_request,
3142 lr_request);
3143 ext4_remove_li_request(elr);
3144 }
3145 mutex_unlock(&ext4_li_info->li_list_mtx);
3146 }
3147
3148 static int ext4_run_lazyinit_thread(void)
3149 {
3150 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3151 ext4_li_info, "ext4lazyinit");
3152 if (IS_ERR(ext4_lazyinit_task)) {
3153 int err = PTR_ERR(ext4_lazyinit_task);
3154 ext4_clear_request_list();
3155 kfree(ext4_li_info);
3156 ext4_li_info = NULL;
3157 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3158 "initialization thread\n",
3159 err);
3160 return err;
3161 }
3162 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3163 return 0;
3164 }
3165
3166 /*
3167 * Check whether it make sense to run itable init. thread or not.
3168 * If there is at least one uninitialized inode table, return
3169 * corresponding group number, else the loop goes through all
3170 * groups and return total number of groups.
3171 */
3172 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3173 {
3174 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3175 struct ext4_group_desc *gdp = NULL;
3176
3177 if (!ext4_has_group_desc_csum(sb))
3178 return ngroups;
3179
3180 for (group = 0; group < ngroups; group++) {
3181 gdp = ext4_get_group_desc(sb, group, NULL);
3182 if (!gdp)
3183 continue;
3184
3185 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3186 break;
3187 }
3188
3189 return group;
3190 }
3191
3192 static int ext4_li_info_new(void)
3193 {
3194 struct ext4_lazy_init *eli = NULL;
3195
3196 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3197 if (!eli)
3198 return -ENOMEM;
3199
3200 INIT_LIST_HEAD(&eli->li_request_list);
3201 mutex_init(&eli->li_list_mtx);
3202
3203 eli->li_state |= EXT4_LAZYINIT_QUIT;
3204
3205 ext4_li_info = eli;
3206
3207 return 0;
3208 }
3209
3210 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3211 ext4_group_t start)
3212 {
3213 struct ext4_sb_info *sbi = EXT4_SB(sb);
3214 struct ext4_li_request *elr;
3215
3216 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3217 if (!elr)
3218 return NULL;
3219
3220 elr->lr_super = sb;
3221 elr->lr_sbi = sbi;
3222 elr->lr_next_group = start;
3223
3224 /*
3225 * Randomize first schedule time of the request to
3226 * spread the inode table initialization requests
3227 * better.
3228 */
3229 elr->lr_next_sched = jiffies + (prandom_u32() %
3230 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3231 return elr;
3232 }
3233
3234 int ext4_register_li_request(struct super_block *sb,
3235 ext4_group_t first_not_zeroed)
3236 {
3237 struct ext4_sb_info *sbi = EXT4_SB(sb);
3238 struct ext4_li_request *elr = NULL;
3239 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3240 int ret = 0;
3241
3242 mutex_lock(&ext4_li_mtx);
3243 if (sbi->s_li_request != NULL) {
3244 /*
3245 * Reset timeout so it can be computed again, because
3246 * s_li_wait_mult might have changed.
3247 */
3248 sbi->s_li_request->lr_timeout = 0;
3249 goto out;
3250 }
3251
3252 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3253 !test_opt(sb, INIT_INODE_TABLE))
3254 goto out;
3255
3256 elr = ext4_li_request_new(sb, first_not_zeroed);
3257 if (!elr) {
3258 ret = -ENOMEM;
3259 goto out;
3260 }
3261
3262 if (NULL == ext4_li_info) {
3263 ret = ext4_li_info_new();
3264 if (ret)
3265 goto out;
3266 }
3267
3268 mutex_lock(&ext4_li_info->li_list_mtx);
3269 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3270 mutex_unlock(&ext4_li_info->li_list_mtx);
3271
3272 sbi->s_li_request = elr;
3273 /*
3274 * set elr to NULL here since it has been inserted to
3275 * the request_list and the removal and free of it is
3276 * handled by ext4_clear_request_list from now on.
3277 */
3278 elr = NULL;
3279
3280 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3281 ret = ext4_run_lazyinit_thread();
3282 if (ret)
3283 goto out;
3284 }
3285 out:
3286 mutex_unlock(&ext4_li_mtx);
3287 if (ret)
3288 kfree(elr);
3289 return ret;
3290 }
3291
3292 /*
3293 * We do not need to lock anything since this is called on
3294 * module unload.
3295 */
3296 static void ext4_destroy_lazyinit_thread(void)
3297 {
3298 /*
3299 * If thread exited earlier
3300 * there's nothing to be done.
3301 */
3302 if (!ext4_li_info || !ext4_lazyinit_task)
3303 return;
3304
3305 kthread_stop(ext4_lazyinit_task);
3306 }
3307
3308 static int set_journal_csum_feature_set(struct super_block *sb)
3309 {
3310 int ret = 1;
3311 int compat, incompat;
3312 struct ext4_sb_info *sbi = EXT4_SB(sb);
3313
3314 if (ext4_has_metadata_csum(sb)) {
3315 /* journal checksum v3 */
3316 compat = 0;
3317 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3318 } else {
3319 /* journal checksum v1 */
3320 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3321 incompat = 0;
3322 }
3323
3324 jbd2_journal_clear_features(sbi->s_journal,
3325 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3326 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3327 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3328 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3329 ret = jbd2_journal_set_features(sbi->s_journal,
3330 compat, 0,
3331 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3332 incompat);
3333 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3334 ret = jbd2_journal_set_features(sbi->s_journal,
3335 compat, 0,
3336 incompat);
3337 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3338 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3339 } else {
3340 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3341 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3342 }
3343
3344 return ret;
3345 }
3346
3347 /*
3348 * Note: calculating the overhead so we can be compatible with
3349 * historical BSD practice is quite difficult in the face of
3350 * clusters/bigalloc. This is because multiple metadata blocks from
3351 * different block group can end up in the same allocation cluster.
3352 * Calculating the exact overhead in the face of clustered allocation
3353 * requires either O(all block bitmaps) in memory or O(number of block
3354 * groups**2) in time. We will still calculate the superblock for
3355 * older file systems --- and if we come across with a bigalloc file
3356 * system with zero in s_overhead_clusters the estimate will be close to
3357 * correct especially for very large cluster sizes --- but for newer
3358 * file systems, it's better to calculate this figure once at mkfs
3359 * time, and store it in the superblock. If the superblock value is
3360 * present (even for non-bigalloc file systems), we will use it.
3361 */
3362 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3363 char *buf)
3364 {
3365 struct ext4_sb_info *sbi = EXT4_SB(sb);
3366 struct ext4_group_desc *gdp;
3367 ext4_fsblk_t first_block, last_block, b;
3368 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3369 int s, j, count = 0;
3370
3371 if (!ext4_has_feature_bigalloc(sb))
3372 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3373 sbi->s_itb_per_group + 2);
3374
3375 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3376 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3377 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3378 for (i = 0; i < ngroups; i++) {
3379 gdp = ext4_get_group_desc(sb, i, NULL);
3380 b = ext4_block_bitmap(sb, gdp);
3381 if (b >= first_block && b <= last_block) {
3382 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3383 count++;
3384 }
3385 b = ext4_inode_bitmap(sb, gdp);
3386 if (b >= first_block && b <= last_block) {
3387 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3388 count++;
3389 }
3390 b = ext4_inode_table(sb, gdp);
3391 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3392 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3393 int c = EXT4_B2C(sbi, b - first_block);
3394 ext4_set_bit(c, buf);
3395 count++;
3396 }
3397 if (i != grp)
3398 continue;
3399 s = 0;
3400 if (ext4_bg_has_super(sb, grp)) {
3401 ext4_set_bit(s++, buf);
3402 count++;
3403 }
3404 j = ext4_bg_num_gdb(sb, grp);
3405 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3406 ext4_error(sb, "Invalid number of block group "
3407 "descriptor blocks: %d", j);
3408 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3409 }
3410 count += j;
3411 for (; j > 0; j--)
3412 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3413 }
3414 if (!count)
3415 return 0;
3416 return EXT4_CLUSTERS_PER_GROUP(sb) -
3417 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3418 }
3419
3420 /*
3421 * Compute the overhead and stash it in sbi->s_overhead
3422 */
3423 int ext4_calculate_overhead(struct super_block *sb)
3424 {
3425 struct ext4_sb_info *sbi = EXT4_SB(sb);
3426 struct ext4_super_block *es = sbi->s_es;
3427 struct inode *j_inode;
3428 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3429 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3430 ext4_fsblk_t overhead = 0;
3431 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3432
3433 if (!buf)
3434 return -ENOMEM;
3435
3436 /*
3437 * Compute the overhead (FS structures). This is constant
3438 * for a given filesystem unless the number of block groups
3439 * changes so we cache the previous value until it does.
3440 */
3441
3442 /*
3443 * All of the blocks before first_data_block are overhead
3444 */
3445 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3446
3447 /*
3448 * Add the overhead found in each block group
3449 */
3450 for (i = 0; i < ngroups; i++) {
3451 int blks;
3452
3453 blks = count_overhead(sb, i, buf);
3454 overhead += blks;
3455 if (blks)
3456 memset(buf, 0, PAGE_SIZE);
3457 cond_resched();
3458 }
3459
3460 /*
3461 * Add the internal journal blocks whether the journal has been
3462 * loaded or not
3463 */
3464 if (sbi->s_journal && !sbi->journal_bdev)
3465 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3466 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3467 j_inode = ext4_get_journal_inode(sb, j_inum);
3468 if (j_inode) {
3469 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3470 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3471 iput(j_inode);
3472 } else {
3473 ext4_msg(sb, KERN_ERR, "can't get journal size");
3474 }
3475 }
3476 sbi->s_overhead = overhead;
3477 smp_wmb();
3478 free_page((unsigned long) buf);
3479 return 0;
3480 }
3481
3482 static void ext4_clamp_want_extra_isize(struct super_block *sb)
3483 {
3484 struct ext4_sb_info *sbi = EXT4_SB(sb);
3485 struct ext4_super_block *es = sbi->s_es;
3486
3487 /* determine the minimum size of new large inodes, if present */
3488 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
3489 sbi->s_want_extra_isize == 0) {
3490 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3491 EXT4_GOOD_OLD_INODE_SIZE;
3492 if (ext4_has_feature_extra_isize(sb)) {
3493 if (sbi->s_want_extra_isize <
3494 le16_to_cpu(es->s_want_extra_isize))
3495 sbi->s_want_extra_isize =
3496 le16_to_cpu(es->s_want_extra_isize);
3497 if (sbi->s_want_extra_isize <
3498 le16_to_cpu(es->s_min_extra_isize))
3499 sbi->s_want_extra_isize =
3500 le16_to_cpu(es->s_min_extra_isize);
3501 }
3502 }
3503 /* Check if enough inode space is available */
3504 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3505 sbi->s_inode_size) {
3506 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3507 EXT4_GOOD_OLD_INODE_SIZE;
3508 ext4_msg(sb, KERN_INFO,
3509 "required extra inode space not available");
3510 }
3511 }
3512
3513 static void ext4_set_resv_clusters(struct super_block *sb)
3514 {
3515 ext4_fsblk_t resv_clusters;
3516 struct ext4_sb_info *sbi = EXT4_SB(sb);
3517
3518 /*
3519 * There's no need to reserve anything when we aren't using extents.
3520 * The space estimates are exact, there are no unwritten extents,
3521 * hole punching doesn't need new metadata... This is needed especially
3522 * to keep ext2/3 backward compatibility.
3523 */
3524 if (!ext4_has_feature_extents(sb))
3525 return;
3526 /*
3527 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3528 * This should cover the situations where we can not afford to run
3529 * out of space like for example punch hole, or converting
3530 * unwritten extents in delalloc path. In most cases such
3531 * allocation would require 1, or 2 blocks, higher numbers are
3532 * very rare.
3533 */
3534 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3535 sbi->s_cluster_bits);
3536
3537 do_div(resv_clusters, 50);
3538 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3539
3540 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3541 }
3542
3543 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3544 {
3545 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3546 char *orig_data = kstrdup(data, GFP_KERNEL);
3547 struct buffer_head *bh;
3548 struct ext4_super_block *es = NULL;
3549 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3550 ext4_fsblk_t block;
3551 ext4_fsblk_t sb_block = get_sb_block(&data);
3552 ext4_fsblk_t logical_sb_block;
3553 unsigned long offset = 0;
3554 unsigned long journal_devnum = 0;
3555 unsigned long def_mount_opts;
3556 struct inode *root;
3557 const char *descr;
3558 int ret = -ENOMEM;
3559 int blocksize, clustersize;
3560 unsigned int db_count;
3561 unsigned int i;
3562 int needs_recovery, has_huge_files, has_bigalloc;
3563 __u64 blocks_count;
3564 int err = 0;
3565 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3566 ext4_group_t first_not_zeroed;
3567
3568 if ((data && !orig_data) || !sbi)
3569 goto out_free_base;
3570
3571 if (!userns_mounts && !capable(CAP_SYS_ADMIN)) {
3572 ret = -EPERM;
3573 goto out_free_base;
3574 }
3575
3576 sbi->s_daxdev = dax_dev;
3577 sbi->s_blockgroup_lock =
3578 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3579 if (!sbi->s_blockgroup_lock)
3580 goto out_free_base;
3581
3582 sb->s_fs_info = sbi;
3583 sbi->s_sb = sb;
3584 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3585 sbi->s_sb_block = sb_block;
3586 if (sb->s_bdev->bd_part)
3587 sbi->s_sectors_written_start =
3588 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3589
3590 /* Cleanup superblock name */
3591 strreplace(sb->s_id, '/', '!');
3592
3593 /* -EINVAL is default */
3594 ret = -EINVAL;
3595 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3596 if (!blocksize) {
3597 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3598 goto out_fail;
3599 }
3600
3601 /*
3602 * The ext4 superblock will not be buffer aligned for other than 1kB
3603 * block sizes. We need to calculate the offset from buffer start.
3604 */
3605 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3606 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3607 offset = do_div(logical_sb_block, blocksize);
3608 } else {
3609 logical_sb_block = sb_block;
3610 }
3611
3612 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3613 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3614 goto out_fail;
3615 }
3616 /*
3617 * Note: s_es must be initialized as soon as possible because
3618 * some ext4 macro-instructions depend on its value
3619 */
3620 es = (struct ext4_super_block *) (bh->b_data + offset);
3621 sbi->s_es = es;
3622 sb->s_magic = le16_to_cpu(es->s_magic);
3623 if (sb->s_magic != EXT4_SUPER_MAGIC)
3624 goto cantfind_ext4;
3625 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3626
3627 /* Warn if metadata_csum and gdt_csum are both set. */
3628 if (ext4_has_feature_metadata_csum(sb) &&
3629 ext4_has_feature_gdt_csum(sb))
3630 ext4_warning(sb, "metadata_csum and uninit_bg are "
3631 "redundant flags; please run fsck.");
3632
3633 /* Check for a known checksum algorithm */
3634 if (!ext4_verify_csum_type(sb, es)) {
3635 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3636 "unknown checksum algorithm.");
3637 silent = 1;
3638 goto cantfind_ext4;
3639 }
3640
3641 /* Load the checksum driver */
3642 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3643 if (IS_ERR(sbi->s_chksum_driver)) {
3644 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3645 ret = PTR_ERR(sbi->s_chksum_driver);
3646 sbi->s_chksum_driver = NULL;
3647 goto failed_mount;
3648 }
3649
3650 /* Check superblock checksum */
3651 if (!ext4_superblock_csum_verify(sb, es)) {
3652 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3653 "invalid superblock checksum. Run e2fsck?");
3654 silent = 1;
3655 ret = -EFSBADCRC;
3656 goto cantfind_ext4;
3657 }
3658
3659 /* Precompute checksum seed for all metadata */
3660 if (ext4_has_feature_csum_seed(sb))
3661 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3662 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3663 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3664 sizeof(es->s_uuid));
3665
3666 /* Set defaults before we parse the mount options */
3667 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3668 set_opt(sb, INIT_INODE_TABLE);
3669 if (def_mount_opts & EXT4_DEFM_DEBUG)
3670 set_opt(sb, DEBUG);
3671 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3672 set_opt(sb, GRPID);
3673 if (def_mount_opts & EXT4_DEFM_UID16)
3674 set_opt(sb, NO_UID32);
3675 /* xattr user namespace & acls are now defaulted on */
3676 set_opt(sb, XATTR_USER);
3677 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3678 set_opt(sb, POSIX_ACL);
3679 #endif
3680 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3681 if (ext4_has_metadata_csum(sb))
3682 set_opt(sb, JOURNAL_CHECKSUM);
3683
3684 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3685 set_opt(sb, JOURNAL_DATA);
3686 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3687 set_opt(sb, ORDERED_DATA);
3688 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3689 set_opt(sb, WRITEBACK_DATA);
3690
3691 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) {
3692 if (!capable(CAP_SYS_ADMIN))
3693 goto failed_mount;
3694 set_opt(sb, ERRORS_PANIC);
3695 } else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) {
3696 set_opt(sb, ERRORS_CONT);
3697 } else {
3698 set_opt(sb, ERRORS_RO);
3699 }
3700 /* block_validity enabled by default; disable with noblock_validity */
3701 set_opt(sb, BLOCK_VALIDITY);
3702 if (def_mount_opts & EXT4_DEFM_DISCARD)
3703 set_opt(sb, DISCARD);
3704
3705 sbi->s_resuid = make_kuid(sb->s_user_ns, le16_to_cpu(es->s_def_resuid));
3706 if (!uid_valid(sbi->s_resuid))
3707 sbi->s_resuid = make_kuid(sb->s_user_ns, EXT4_DEF_RESUID);
3708 sbi->s_resgid = make_kgid(sb->s_user_ns, le16_to_cpu(es->s_def_resgid));
3709 if (!gid_valid(sbi->s_resgid))
3710 sbi->s_resgid = make_kgid(sb->s_user_ns, EXT4_DEF_RESGID);
3711 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3712 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3713 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3714
3715 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3716 set_opt(sb, BARRIER);
3717
3718 /*
3719 * enable delayed allocation by default
3720 * Use -o nodelalloc to turn it off
3721 */
3722 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3723 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3724 set_opt(sb, DELALLOC);
3725
3726 /*
3727 * set default s_li_wait_mult for lazyinit, for the case there is
3728 * no mount option specified.
3729 */
3730 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3731
3732 if (sbi->s_es->s_mount_opts[0]) {
3733 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3734 sizeof(sbi->s_es->s_mount_opts),
3735 GFP_KERNEL);
3736 if (!s_mount_opts)
3737 goto failed_mount;
3738 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3739 &journal_ioprio, 0)) {
3740 ext4_msg(sb, KERN_WARNING,
3741 "failed to parse options in superblock: %s",
3742 s_mount_opts);
3743 }
3744 kfree(s_mount_opts);
3745 }
3746 sbi->s_def_mount_opt = sbi->s_mount_opt;
3747 if (!parse_options((char *) data, sb, &journal_devnum,
3748 &journal_ioprio, 0))
3749 goto failed_mount;
3750
3751 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3752 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3753 "with data=journal disables delayed "
3754 "allocation and O_DIRECT support!\n");
3755 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3756 ext4_msg(sb, KERN_ERR, "can't mount with "
3757 "both data=journal and delalloc");
3758 goto failed_mount;
3759 }
3760 if (test_opt(sb, DIOREAD_NOLOCK)) {
3761 ext4_msg(sb, KERN_ERR, "can't mount with "
3762 "both data=journal and dioread_nolock");
3763 goto failed_mount;
3764 }
3765 if (test_opt(sb, DAX)) {
3766 ext4_msg(sb, KERN_ERR, "can't mount with "
3767 "both data=journal and dax");
3768 goto failed_mount;
3769 }
3770 if (ext4_has_feature_encrypt(sb)) {
3771 ext4_msg(sb, KERN_WARNING,
3772 "encrypted files will use data=ordered "
3773 "instead of data journaling mode");
3774 }
3775 if (test_opt(sb, DELALLOC))
3776 clear_opt(sb, DELALLOC);
3777 } else {
3778 sb->s_iflags |= SB_I_CGROUPWB;
3779 }
3780
3781 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3782 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
3783
3784 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3785 (ext4_has_compat_features(sb) ||
3786 ext4_has_ro_compat_features(sb) ||
3787 ext4_has_incompat_features(sb)))
3788 ext4_msg(sb, KERN_WARNING,
3789 "feature flags set on rev 0 fs, "
3790 "running e2fsck is recommended");
3791
3792 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3793 set_opt2(sb, HURD_COMPAT);
3794 if (ext4_has_feature_64bit(sb)) {
3795 ext4_msg(sb, KERN_ERR,
3796 "The Hurd can't support 64-bit file systems");
3797 goto failed_mount;
3798 }
3799
3800 /*
3801 * ea_inode feature uses l_i_version field which is not
3802 * available in HURD_COMPAT mode.
3803 */
3804 if (ext4_has_feature_ea_inode(sb)) {
3805 ext4_msg(sb, KERN_ERR,
3806 "ea_inode feature is not supported for Hurd");
3807 goto failed_mount;
3808 }
3809 }
3810
3811 if (IS_EXT2_SB(sb)) {
3812 if (ext2_feature_set_ok(sb))
3813 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3814 "using the ext4 subsystem");
3815 else {
3816 /*
3817 * If we're probing be silent, if this looks like
3818 * it's actually an ext[34] filesystem.
3819 */
3820 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3821 goto failed_mount;
3822 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3823 "to feature incompatibilities");
3824 goto failed_mount;
3825 }
3826 }
3827
3828 if (IS_EXT3_SB(sb)) {
3829 if (ext3_feature_set_ok(sb))
3830 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3831 "using the ext4 subsystem");
3832 else {
3833 /*
3834 * If we're probing be silent, if this looks like
3835 * it's actually an ext4 filesystem.
3836 */
3837 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3838 goto failed_mount;
3839 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3840 "to feature incompatibilities");
3841 goto failed_mount;
3842 }
3843 }
3844
3845 /*
3846 * Check feature flags regardless of the revision level, since we
3847 * previously didn't change the revision level when setting the flags,
3848 * so there is a chance incompat flags are set on a rev 0 filesystem.
3849 */
3850 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
3851 goto failed_mount;
3852
3853 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3854 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3855 blocksize > EXT4_MAX_BLOCK_SIZE) {
3856 ext4_msg(sb, KERN_ERR,
3857 "Unsupported filesystem blocksize %d (%d log_block_size)",
3858 blocksize, le32_to_cpu(es->s_log_block_size));
3859 goto failed_mount;
3860 }
3861 if (le32_to_cpu(es->s_log_block_size) >
3862 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3863 ext4_msg(sb, KERN_ERR,
3864 "Invalid log block size: %u",
3865 le32_to_cpu(es->s_log_block_size));
3866 goto failed_mount;
3867 }
3868 if (le32_to_cpu(es->s_log_cluster_size) >
3869 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3870 ext4_msg(sb, KERN_ERR,
3871 "Invalid log cluster size: %u",
3872 le32_to_cpu(es->s_log_cluster_size));
3873 goto failed_mount;
3874 }
3875
3876 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3877 ext4_msg(sb, KERN_ERR,
3878 "Number of reserved GDT blocks insanely large: %d",
3879 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3880 goto failed_mount;
3881 }
3882
3883 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3884 if (ext4_has_feature_inline_data(sb)) {
3885 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
3886 " that may contain inline data");
3887 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
3888 }
3889 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
3890 ext4_msg(sb, KERN_ERR,
3891 "DAX unsupported by block device. Turning off DAX.");
3892 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
3893 }
3894 }
3895
3896 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3897 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3898 es->s_encryption_level);
3899 goto failed_mount;
3900 }
3901
3902 if (sb->s_blocksize != blocksize) {
3903 /* Validate the filesystem blocksize */
3904 if (!sb_set_blocksize(sb, blocksize)) {
3905 ext4_msg(sb, KERN_ERR, "bad block size %d",
3906 blocksize);
3907 goto failed_mount;
3908 }
3909
3910 brelse(bh);
3911 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3912 offset = do_div(logical_sb_block, blocksize);
3913 bh = sb_bread_unmovable(sb, logical_sb_block);
3914 if (!bh) {
3915 ext4_msg(sb, KERN_ERR,
3916 "Can't read superblock on 2nd try");
3917 goto failed_mount;
3918 }
3919 es = (struct ext4_super_block *)(bh->b_data + offset);
3920 sbi->s_es = es;
3921 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3922 ext4_msg(sb, KERN_ERR,
3923 "Magic mismatch, very weird!");
3924 goto failed_mount;
3925 }
3926 }
3927
3928 has_huge_files = ext4_has_feature_huge_file(sb);
3929 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3930 has_huge_files);
3931 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3932
3933 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3934 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3935 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3936 } else {
3937 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3938 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3939 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3940 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3941 sbi->s_first_ino);
3942 goto failed_mount;
3943 }
3944 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3945 (!is_power_of_2(sbi->s_inode_size)) ||
3946 (sbi->s_inode_size > blocksize)) {
3947 ext4_msg(sb, KERN_ERR,
3948 "unsupported inode size: %d",
3949 sbi->s_inode_size);
3950 goto failed_mount;
3951 }
3952 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3953 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3954 }
3955
3956 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3957 if (ext4_has_feature_64bit(sb)) {
3958 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3959 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3960 !is_power_of_2(sbi->s_desc_size)) {
3961 ext4_msg(sb, KERN_ERR,
3962 "unsupported descriptor size %lu",
3963 sbi->s_desc_size);
3964 goto failed_mount;
3965 }
3966 } else
3967 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3968
3969 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3970 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3971
3972 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3973 if (sbi->s_inodes_per_block == 0)
3974 goto cantfind_ext4;
3975 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3976 sbi->s_inodes_per_group > blocksize * 8) {
3977 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3978 sbi->s_blocks_per_group);
3979 goto failed_mount;
3980 }
3981 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3982 sbi->s_inodes_per_block;
3983 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3984 sbi->s_sbh = bh;
3985 sbi->s_mount_state = le16_to_cpu(es->s_state);
3986 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3987 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3988
3989 for (i = 0; i < 4; i++)
3990 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3991 sbi->s_def_hash_version = es->s_def_hash_version;
3992 if (ext4_has_feature_dir_index(sb)) {
3993 i = le32_to_cpu(es->s_flags);
3994 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3995 sbi->s_hash_unsigned = 3;
3996 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3997 #ifdef __CHAR_UNSIGNED__
3998 if (!sb_rdonly(sb))
3999 es->s_flags |=
4000 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4001 sbi->s_hash_unsigned = 3;
4002 #else
4003 if (!sb_rdonly(sb))
4004 es->s_flags |=
4005 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4006 #endif
4007 }
4008 }
4009
4010 /* Handle clustersize */
4011 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4012 has_bigalloc = ext4_has_feature_bigalloc(sb);
4013 if (has_bigalloc) {
4014 if (clustersize < blocksize) {
4015 ext4_msg(sb, KERN_ERR,
4016 "cluster size (%d) smaller than "
4017 "block size (%d)", clustersize, blocksize);
4018 goto failed_mount;
4019 }
4020 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4021 le32_to_cpu(es->s_log_block_size);
4022 sbi->s_clusters_per_group =
4023 le32_to_cpu(es->s_clusters_per_group);
4024 if (sbi->s_clusters_per_group > blocksize * 8) {
4025 ext4_msg(sb, KERN_ERR,
4026 "#clusters per group too big: %lu",
4027 sbi->s_clusters_per_group);
4028 goto failed_mount;
4029 }
4030 if (sbi->s_blocks_per_group !=
4031 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4032 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4033 "clusters per group (%lu) inconsistent",
4034 sbi->s_blocks_per_group,
4035 sbi->s_clusters_per_group);
4036 goto failed_mount;
4037 }
4038 } else {
4039 if (clustersize != blocksize) {
4040 ext4_msg(sb, KERN_ERR,
4041 "fragment/cluster size (%d) != "
4042 "block size (%d)", clustersize, blocksize);
4043 goto failed_mount;
4044 }
4045 if (sbi->s_blocks_per_group > blocksize * 8) {
4046 ext4_msg(sb, KERN_ERR,
4047 "#blocks per group too big: %lu",
4048 sbi->s_blocks_per_group);
4049 goto failed_mount;
4050 }
4051 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4052 sbi->s_cluster_bits = 0;
4053 }
4054 sbi->s_cluster_ratio = clustersize / blocksize;
4055
4056 /* Do we have standard group size of clustersize * 8 blocks ? */
4057 if (sbi->s_blocks_per_group == clustersize << 3)
4058 set_opt2(sb, STD_GROUP_SIZE);
4059
4060 /*
4061 * Test whether we have more sectors than will fit in sector_t,
4062 * and whether the max offset is addressable by the page cache.
4063 */
4064 err = generic_check_addressable(sb->s_blocksize_bits,
4065 ext4_blocks_count(es));
4066 if (err) {
4067 ext4_msg(sb, KERN_ERR, "filesystem"
4068 " too large to mount safely on this system");
4069 if (sizeof(sector_t) < 8)
4070 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
4071 goto failed_mount;
4072 }
4073
4074 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4075 goto cantfind_ext4;
4076
4077 /* check blocks count against device size */
4078 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4079 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4080 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4081 "exceeds size of device (%llu blocks)",
4082 ext4_blocks_count(es), blocks_count);
4083 goto failed_mount;
4084 }
4085
4086 /*
4087 * It makes no sense for the first data block to be beyond the end
4088 * of the filesystem.
4089 */
4090 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4091 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4092 "block %u is beyond end of filesystem (%llu)",
4093 le32_to_cpu(es->s_first_data_block),
4094 ext4_blocks_count(es));
4095 goto failed_mount;
4096 }
4097 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4098 (sbi->s_cluster_ratio == 1)) {
4099 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4100 "block is 0 with a 1k block and cluster size");
4101 goto failed_mount;
4102 }
4103
4104 blocks_count = (ext4_blocks_count(es) -
4105 le32_to_cpu(es->s_first_data_block) +
4106 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4107 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4108 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4109 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4110 "(block count %llu, first data block %u, "
4111 "blocks per group %lu)", sbi->s_groups_count,
4112 ext4_blocks_count(es),
4113 le32_to_cpu(es->s_first_data_block),
4114 EXT4_BLOCKS_PER_GROUP(sb));
4115 goto failed_mount;
4116 }
4117 sbi->s_groups_count = blocks_count;
4118 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4119 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4120 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4121 le32_to_cpu(es->s_inodes_count)) {
4122 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4123 le32_to_cpu(es->s_inodes_count),
4124 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4125 ret = -EINVAL;
4126 goto failed_mount;
4127 }
4128 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4129 EXT4_DESC_PER_BLOCK(sb);
4130 if (ext4_has_feature_meta_bg(sb)) {
4131 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4132 ext4_msg(sb, KERN_WARNING,
4133 "first meta block group too large: %u "
4134 "(group descriptor block count %u)",
4135 le32_to_cpu(es->s_first_meta_bg), db_count);
4136 goto failed_mount;
4137 }
4138 }
4139 sbi->s_group_desc = kvmalloc(db_count *
4140 sizeof(struct buffer_head *),
4141 GFP_KERNEL);
4142 if (sbi->s_group_desc == NULL) {
4143 ext4_msg(sb, KERN_ERR, "not enough memory");
4144 ret = -ENOMEM;
4145 goto failed_mount;
4146 }
4147
4148 bgl_lock_init(sbi->s_blockgroup_lock);
4149
4150 /* Pre-read the descriptors into the buffer cache */
4151 for (i = 0; i < db_count; i++) {
4152 block = descriptor_loc(sb, logical_sb_block, i);
4153 sb_breadahead(sb, block);
4154 }
4155
4156 for (i = 0; i < db_count; i++) {
4157 block = descriptor_loc(sb, logical_sb_block, i);
4158 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
4159 if (!sbi->s_group_desc[i]) {
4160 ext4_msg(sb, KERN_ERR,
4161 "can't read group descriptor %d", i);
4162 db_count = i;
4163 goto failed_mount2;
4164 }
4165 }
4166 sbi->s_gdb_count = db_count;
4167 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4168 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4169 ret = -EFSCORRUPTED;
4170 goto failed_mount2;
4171 }
4172
4173 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4174
4175 /* Register extent status tree shrinker */
4176 if (ext4_es_register_shrinker(sbi))
4177 goto failed_mount3;
4178
4179 sbi->s_stripe = ext4_get_stripe_size(sbi);
4180 sbi->s_extent_max_zeroout_kb = 32;
4181
4182 /*
4183 * set up enough so that it can read an inode
4184 */
4185 sb->s_op = &ext4_sops;
4186 sb->s_export_op = &ext4_export_ops;
4187 sb->s_xattr = ext4_xattr_handlers;
4188 #ifdef CONFIG_EXT4_FS_ENCRYPTION
4189 sb->s_cop = &ext4_cryptops;
4190 #endif
4191 #ifdef CONFIG_QUOTA
4192 sb->dq_op = &ext4_quota_operations;
4193 if (ext4_has_feature_quota(sb))
4194 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4195 else
4196 sb->s_qcop = &ext4_qctl_operations;
4197 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4198 #endif
4199 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4200
4201 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4202 mutex_init(&sbi->s_orphan_lock);
4203
4204 sb->s_root = NULL;
4205
4206 needs_recovery = (es->s_last_orphan != 0 ||
4207 ext4_has_feature_journal_needs_recovery(sb));
4208
4209 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4210 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4211 goto failed_mount3a;
4212
4213 /*
4214 * The first inode we look at is the journal inode. Don't try
4215 * root first: it may be modified in the journal!
4216 */
4217 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4218 err = ext4_load_journal(sb, es, journal_devnum);
4219 if (err)
4220 goto failed_mount3a;
4221 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4222 ext4_has_feature_journal_needs_recovery(sb)) {
4223 ext4_msg(sb, KERN_ERR, "required journal recovery "
4224 "suppressed and not mounted read-only");
4225 goto failed_mount_wq;
4226 } else {
4227 /* Nojournal mode, all journal mount options are illegal */
4228 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4229 ext4_msg(sb, KERN_ERR, "can't mount with "
4230 "journal_checksum, fs mounted w/o journal");
4231 goto failed_mount_wq;
4232 }
4233 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4234 ext4_msg(sb, KERN_ERR, "can't mount with "
4235 "journal_async_commit, fs mounted w/o journal");
4236 goto failed_mount_wq;
4237 }
4238 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4239 ext4_msg(sb, KERN_ERR, "can't mount with "
4240 "commit=%lu, fs mounted w/o journal",
4241 sbi->s_commit_interval / HZ);
4242 goto failed_mount_wq;
4243 }
4244 if (EXT4_MOUNT_DATA_FLAGS &
4245 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4246 ext4_msg(sb, KERN_ERR, "can't mount with "
4247 "data=, fs mounted w/o journal");
4248 goto failed_mount_wq;
4249 }
4250 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4251 clear_opt(sb, JOURNAL_CHECKSUM);
4252 clear_opt(sb, DATA_FLAGS);
4253 sbi->s_journal = NULL;
4254 needs_recovery = 0;
4255 goto no_journal;
4256 }
4257
4258 if (ext4_has_feature_64bit(sb) &&
4259 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4260 JBD2_FEATURE_INCOMPAT_64BIT)) {
4261 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4262 goto failed_mount_wq;
4263 }
4264
4265 if (!set_journal_csum_feature_set(sb)) {
4266 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4267 "feature set");
4268 goto failed_mount_wq;
4269 }
4270
4271 /* We have now updated the journal if required, so we can
4272 * validate the data journaling mode. */
4273 switch (test_opt(sb, DATA_FLAGS)) {
4274 case 0:
4275 /* No mode set, assume a default based on the journal
4276 * capabilities: ORDERED_DATA if the journal can
4277 * cope, else JOURNAL_DATA
4278 */
4279 if (jbd2_journal_check_available_features
4280 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4281 set_opt(sb, ORDERED_DATA);
4282 else
4283 set_opt(sb, JOURNAL_DATA);
4284 break;
4285
4286 case EXT4_MOUNT_ORDERED_DATA:
4287 case EXT4_MOUNT_WRITEBACK_DATA:
4288 if (!jbd2_journal_check_available_features
4289 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4290 ext4_msg(sb, KERN_ERR, "Journal does not support "
4291 "requested data journaling mode");
4292 goto failed_mount_wq;
4293 }
4294 default:
4295 break;
4296 }
4297
4298 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4299 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4300 ext4_msg(sb, KERN_ERR, "can't mount with "
4301 "journal_async_commit in data=ordered mode");
4302 goto failed_mount_wq;
4303 }
4304
4305 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4306
4307 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4308
4309 no_journal:
4310 if (!test_opt(sb, NO_MBCACHE)) {
4311 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4312 if (!sbi->s_ea_block_cache) {
4313 ext4_msg(sb, KERN_ERR,
4314 "Failed to create ea_block_cache");
4315 goto failed_mount_wq;
4316 }
4317
4318 if (ext4_has_feature_ea_inode(sb)) {
4319 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4320 if (!sbi->s_ea_inode_cache) {
4321 ext4_msg(sb, KERN_ERR,
4322 "Failed to create ea_inode_cache");
4323 goto failed_mount_wq;
4324 }
4325 }
4326 }
4327
4328 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4329 (blocksize != PAGE_SIZE)) {
4330 ext4_msg(sb, KERN_ERR,
4331 "Unsupported blocksize for fs encryption");
4332 goto failed_mount_wq;
4333 }
4334
4335 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4336 !ext4_has_feature_encrypt(sb)) {
4337 ext4_set_feature_encrypt(sb);
4338 ext4_commit_super(sb, 1);
4339 }
4340
4341 /*
4342 * Get the # of file system overhead blocks from the
4343 * superblock if present.
4344 */
4345 if (es->s_overhead_clusters)
4346 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4347 else {
4348 err = ext4_calculate_overhead(sb);
4349 if (err)
4350 goto failed_mount_wq;
4351 }
4352
4353 /*
4354 * The maximum number of concurrent works can be high and
4355 * concurrency isn't really necessary. Limit it to 1.
4356 */
4357 EXT4_SB(sb)->rsv_conversion_wq =
4358 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4359 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4360 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4361 ret = -ENOMEM;
4362 goto failed_mount4;
4363 }
4364
4365 /*
4366 * The jbd2_journal_load will have done any necessary log recovery,
4367 * so we can safely mount the rest of the filesystem now.
4368 */
4369
4370 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4371 if (IS_ERR(root)) {
4372 ext4_msg(sb, KERN_ERR, "get root inode failed");
4373 ret = PTR_ERR(root);
4374 root = NULL;
4375 goto failed_mount4;
4376 }
4377 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4378 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4379 iput(root);
4380 goto failed_mount4;
4381 }
4382 sb->s_root = d_make_root(root);
4383 if (!sb->s_root) {
4384 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4385 ret = -ENOMEM;
4386 goto failed_mount4;
4387 }
4388
4389 if (ext4_setup_super(sb, es, sb_rdonly(sb)))
4390 sb->s_flags |= SB_RDONLY;
4391
4392 ext4_clamp_want_extra_isize(sb);
4393
4394 ext4_set_resv_clusters(sb);
4395
4396 err = ext4_setup_system_zone(sb);
4397 if (err) {
4398 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4399 "zone (%d)", err);
4400 goto failed_mount4a;
4401 }
4402
4403 ext4_ext_init(sb);
4404 err = ext4_mb_init(sb);
4405 if (err) {
4406 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4407 err);
4408 goto failed_mount5;
4409 }
4410
4411 block = ext4_count_free_clusters(sb);
4412 ext4_free_blocks_count_set(sbi->s_es,
4413 EXT4_C2B(sbi, block));
4414 ext4_superblock_csum_set(sb);
4415 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4416 GFP_KERNEL);
4417 if (!err) {
4418 unsigned long freei = ext4_count_free_inodes(sb);
4419 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4420 ext4_superblock_csum_set(sb);
4421 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4422 GFP_KERNEL);
4423 }
4424 if (!err)
4425 err = percpu_counter_init(&sbi->s_dirs_counter,
4426 ext4_count_dirs(sb), GFP_KERNEL);
4427 if (!err)
4428 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4429 GFP_KERNEL);
4430 if (!err)
4431 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4432
4433 if (err) {
4434 ext4_msg(sb, KERN_ERR, "insufficient memory");
4435 goto failed_mount6;
4436 }
4437
4438 if (ext4_has_feature_flex_bg(sb))
4439 if (!ext4_fill_flex_info(sb)) {
4440 ext4_msg(sb, KERN_ERR,
4441 "unable to initialize "
4442 "flex_bg meta info!");
4443 goto failed_mount6;
4444 }
4445
4446 err = ext4_register_li_request(sb, first_not_zeroed);
4447 if (err)
4448 goto failed_mount6;
4449
4450 err = ext4_register_sysfs(sb);
4451 if (err)
4452 goto failed_mount7;
4453
4454 #ifdef CONFIG_QUOTA
4455 /* Enable quota usage during mount. */
4456 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4457 err = ext4_enable_quotas(sb);
4458 if (err)
4459 goto failed_mount8;
4460 }
4461 #endif /* CONFIG_QUOTA */
4462
4463 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4464 ext4_orphan_cleanup(sb, es);
4465 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4466 if (needs_recovery) {
4467 ext4_msg(sb, KERN_INFO, "recovery complete");
4468 ext4_mark_recovery_complete(sb, es);
4469 }
4470 if (EXT4_SB(sb)->s_journal) {
4471 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4472 descr = " journalled data mode";
4473 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4474 descr = " ordered data mode";
4475 else
4476 descr = " writeback data mode";
4477 } else
4478 descr = "out journal";
4479
4480 if (test_opt(sb, DISCARD)) {
4481 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4482 if (!blk_queue_discard(q))
4483 ext4_msg(sb, KERN_WARNING,
4484 "mounting with \"discard\" option, but "
4485 "the device does not support discard");
4486 }
4487
4488 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4489 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4490 "Opts: %.*s%s%s", descr,
4491 (int) sizeof(sbi->s_es->s_mount_opts),
4492 sbi->s_es->s_mount_opts,
4493 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4494
4495 if (es->s_error_count)
4496 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4497
4498 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4499 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4500 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4501 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4502
4503 kfree(orig_data);
4504 return 0;
4505
4506 cantfind_ext4:
4507 if (!silent)
4508 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4509 goto failed_mount;
4510
4511 #ifdef CONFIG_QUOTA
4512 failed_mount8:
4513 ext4_unregister_sysfs(sb);
4514 #endif
4515 failed_mount7:
4516 ext4_unregister_li_request(sb);
4517 failed_mount6:
4518 ext4_mb_release(sb);
4519 if (sbi->s_flex_groups)
4520 kvfree(sbi->s_flex_groups);
4521 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4522 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4523 percpu_counter_destroy(&sbi->s_dirs_counter);
4524 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4525 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
4526 failed_mount5:
4527 ext4_ext_release(sb);
4528 ext4_release_system_zone(sb);
4529 failed_mount4a:
4530 dput(sb->s_root);
4531 sb->s_root = NULL;
4532 failed_mount4:
4533 ext4_msg(sb, KERN_ERR, "mount failed");
4534 if (EXT4_SB(sb)->rsv_conversion_wq)
4535 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4536 failed_mount_wq:
4537 if (sbi->s_ea_inode_cache) {
4538 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4539 sbi->s_ea_inode_cache = NULL;
4540 }
4541 if (sbi->s_ea_block_cache) {
4542 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4543 sbi->s_ea_block_cache = NULL;
4544 }
4545 if (sbi->s_journal) {
4546 jbd2_journal_destroy(sbi->s_journal);
4547 sbi->s_journal = NULL;
4548 }
4549 failed_mount3a:
4550 ext4_es_unregister_shrinker(sbi);
4551 failed_mount3:
4552 del_timer_sync(&sbi->s_err_report);
4553 if (sbi->s_mmp_tsk)
4554 kthread_stop(sbi->s_mmp_tsk);
4555 failed_mount2:
4556 for (i = 0; i < db_count; i++)
4557 brelse(sbi->s_group_desc[i]);
4558 kvfree(sbi->s_group_desc);
4559 failed_mount:
4560 if (sbi->s_chksum_driver)
4561 crypto_free_shash(sbi->s_chksum_driver);
4562 #ifdef CONFIG_QUOTA
4563 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4564 kfree(sbi->s_qf_names[i]);
4565 #endif
4566 ext4_blkdev_remove(sbi);
4567 brelse(bh);
4568 out_fail:
4569 /* sb->s_user_ns will be put when sb is destroyed */
4570 sb->s_fs_info = NULL;
4571 kfree(sbi->s_blockgroup_lock);
4572 out_free_base:
4573 kfree(sbi);
4574 kfree(orig_data);
4575 fs_put_dax(dax_dev);
4576 return err ? err : ret;
4577 }
4578
4579 /*
4580 * Setup any per-fs journal parameters now. We'll do this both on
4581 * initial mount, once the journal has been initialised but before we've
4582 * done any recovery; and again on any subsequent remount.
4583 */
4584 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4585 {
4586 struct ext4_sb_info *sbi = EXT4_SB(sb);
4587
4588 journal->j_commit_interval = sbi->s_commit_interval;
4589 journal->j_min_batch_time = sbi->s_min_batch_time;
4590 journal->j_max_batch_time = sbi->s_max_batch_time;
4591
4592 write_lock(&journal->j_state_lock);
4593 if (test_opt(sb, BARRIER))
4594 journal->j_flags |= JBD2_BARRIER;
4595 else
4596 journal->j_flags &= ~JBD2_BARRIER;
4597 if (test_opt(sb, DATA_ERR_ABORT))
4598 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4599 else
4600 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4601 write_unlock(&journal->j_state_lock);
4602 }
4603
4604 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4605 unsigned int journal_inum)
4606 {
4607 struct inode *journal_inode;
4608
4609 /*
4610 * Test for the existence of a valid inode on disk. Bad things
4611 * happen if we iget() an unused inode, as the subsequent iput()
4612 * will try to delete it.
4613 */
4614 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
4615 if (IS_ERR(journal_inode)) {
4616 ext4_msg(sb, KERN_ERR, "no journal found");
4617 return NULL;
4618 }
4619 if (!journal_inode->i_nlink) {
4620 make_bad_inode(journal_inode);
4621 iput(journal_inode);
4622 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4623 return NULL;
4624 }
4625
4626 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4627 journal_inode, journal_inode->i_size);
4628 if (!S_ISREG(journal_inode->i_mode)) {
4629 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4630 iput(journal_inode);
4631 return NULL;
4632 }
4633 return journal_inode;
4634 }
4635
4636 static journal_t *ext4_get_journal(struct super_block *sb,
4637 unsigned int journal_inum)
4638 {
4639 struct inode *journal_inode;
4640 journal_t *journal;
4641
4642 BUG_ON(!ext4_has_feature_journal(sb));
4643
4644 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4645 if (!journal_inode)
4646 return NULL;
4647
4648 journal = jbd2_journal_init_inode(journal_inode);
4649 if (!journal) {
4650 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4651 iput(journal_inode);
4652 return NULL;
4653 }
4654 journal->j_private = sb;
4655 ext4_init_journal_params(sb, journal);
4656 return journal;
4657 }
4658
4659 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4660 dev_t j_dev)
4661 {
4662 struct buffer_head *bh;
4663 journal_t *journal;
4664 ext4_fsblk_t start;
4665 ext4_fsblk_t len;
4666 int hblock, blocksize;
4667 ext4_fsblk_t sb_block;
4668 unsigned long offset;
4669 struct ext4_super_block *es;
4670 struct block_device *bdev;
4671
4672 BUG_ON(!ext4_has_feature_journal(sb));
4673
4674 bdev = ext4_blkdev_get(j_dev, sb);
4675 if (bdev == NULL)
4676 return NULL;
4677
4678 blocksize = sb->s_blocksize;
4679 hblock = bdev_logical_block_size(bdev);
4680 if (blocksize < hblock) {
4681 ext4_msg(sb, KERN_ERR,
4682 "blocksize too small for journal device");
4683 goto out_bdev;
4684 }
4685
4686 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4687 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4688 set_blocksize(bdev, blocksize);
4689 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4690 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4691 "external journal");
4692 goto out_bdev;
4693 }
4694
4695 es = (struct ext4_super_block *) (bh->b_data + offset);
4696 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4697 !(le32_to_cpu(es->s_feature_incompat) &
4698 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4699 ext4_msg(sb, KERN_ERR, "external journal has "
4700 "bad superblock");
4701 brelse(bh);
4702 goto out_bdev;
4703 }
4704
4705 if ((le32_to_cpu(es->s_feature_ro_compat) &
4706 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4707 es->s_checksum != ext4_superblock_csum(sb, es)) {
4708 ext4_msg(sb, KERN_ERR, "external journal has "
4709 "corrupt superblock");
4710 brelse(bh);
4711 goto out_bdev;
4712 }
4713
4714 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4715 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4716 brelse(bh);
4717 goto out_bdev;
4718 }
4719
4720 len = ext4_blocks_count(es);
4721 start = sb_block + 1;
4722 brelse(bh); /* we're done with the superblock */
4723
4724 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4725 start, len, blocksize);
4726 if (!journal) {
4727 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4728 goto out_bdev;
4729 }
4730 journal->j_private = sb;
4731 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4732 wait_on_buffer(journal->j_sb_buffer);
4733 if (!buffer_uptodate(journal->j_sb_buffer)) {
4734 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4735 goto out_journal;
4736 }
4737 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4738 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4739 "user (unsupported) - %d",
4740 be32_to_cpu(journal->j_superblock->s_nr_users));
4741 goto out_journal;
4742 }
4743 EXT4_SB(sb)->journal_bdev = bdev;
4744 ext4_init_journal_params(sb, journal);
4745 return journal;
4746
4747 out_journal:
4748 jbd2_journal_destroy(journal);
4749 out_bdev:
4750 ext4_blkdev_put(bdev);
4751 return NULL;
4752 }
4753
4754 static int ext4_load_journal(struct super_block *sb,
4755 struct ext4_super_block *es,
4756 unsigned long journal_devnum)
4757 {
4758 journal_t *journal;
4759 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4760 dev_t journal_dev;
4761 int err = 0;
4762 int really_read_only;
4763
4764 BUG_ON(!ext4_has_feature_journal(sb));
4765
4766 if (journal_devnum &&
4767 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4768 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4769 "numbers have changed");
4770 journal_dev = new_decode_dev(journal_devnum);
4771 } else
4772 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4773
4774 really_read_only = bdev_read_only(sb->s_bdev);
4775
4776 /*
4777 * Are we loading a blank journal or performing recovery after a
4778 * crash? For recovery, we need to check in advance whether we
4779 * can get read-write access to the device.
4780 */
4781 if (ext4_has_feature_journal_needs_recovery(sb)) {
4782 if (sb_rdonly(sb)) {
4783 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4784 "required on readonly filesystem");
4785 if (really_read_only) {
4786 ext4_msg(sb, KERN_ERR, "write access "
4787 "unavailable, cannot proceed "
4788 "(try mounting with noload)");
4789 return -EROFS;
4790 }
4791 ext4_msg(sb, KERN_INFO, "write access will "
4792 "be enabled during recovery");
4793 }
4794 }
4795
4796 if (journal_inum && journal_dev) {
4797 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4798 "and inode journals!");
4799 return -EINVAL;
4800 }
4801
4802 if (journal_inum) {
4803 if (!(journal = ext4_get_journal(sb, journal_inum)))
4804 return -EINVAL;
4805 } else {
4806 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4807 return -EINVAL;
4808 }
4809
4810 if (!(journal->j_flags & JBD2_BARRIER))
4811 ext4_msg(sb, KERN_INFO, "barriers disabled");
4812
4813 if (!ext4_has_feature_journal_needs_recovery(sb))
4814 err = jbd2_journal_wipe(journal, !really_read_only);
4815 if (!err) {
4816 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4817 if (save)
4818 memcpy(save, ((char *) es) +
4819 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4820 err = jbd2_journal_load(journal);
4821 if (save)
4822 memcpy(((char *) es) + EXT4_S_ERR_START,
4823 save, EXT4_S_ERR_LEN);
4824 kfree(save);
4825 }
4826
4827 if (err) {
4828 ext4_msg(sb, KERN_ERR, "error loading journal");
4829 jbd2_journal_destroy(journal);
4830 return err;
4831 }
4832
4833 EXT4_SB(sb)->s_journal = journal;
4834 ext4_clear_journal_err(sb, es);
4835
4836 if (!really_read_only && journal_devnum &&
4837 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4838 es->s_journal_dev = cpu_to_le32(journal_devnum);
4839
4840 /* Make sure we flush the recovery flag to disk. */
4841 ext4_commit_super(sb, 1);
4842 }
4843
4844 return 0;
4845 }
4846
4847 static int ext4_commit_super(struct super_block *sb, int sync)
4848 {
4849 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4850 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4851 int error = 0;
4852
4853 if (!sbh || block_device_ejected(sb))
4854 return error;
4855
4856 /*
4857 * The superblock bh should be mapped, but it might not be if the
4858 * device was hot-removed. Not much we can do but fail the I/O.
4859 */
4860 if (!buffer_mapped(sbh))
4861 return error;
4862
4863 /*
4864 * If the file system is mounted read-only, don't update the
4865 * superblock write time. This avoids updating the superblock
4866 * write time when we are mounting the root file system
4867 * read/only but we need to replay the journal; at that point,
4868 * for people who are east of GMT and who make their clock
4869 * tick in localtime for Windows bug-for-bug compatibility,
4870 * the clock is set in the future, and this will cause e2fsck
4871 * to complain and force a full file system check.
4872 */
4873 if (!(sb->s_flags & SB_RDONLY))
4874 es->s_wtime = cpu_to_le32(get_seconds());
4875 if (sb->s_bdev->bd_part)
4876 es->s_kbytes_written =
4877 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4878 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4879 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4880 else
4881 es->s_kbytes_written =
4882 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4883 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4884 ext4_free_blocks_count_set(es,
4885 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4886 &EXT4_SB(sb)->s_freeclusters_counter)));
4887 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4888 es->s_free_inodes_count =
4889 cpu_to_le32(percpu_counter_sum_positive(
4890 &EXT4_SB(sb)->s_freeinodes_counter));
4891 BUFFER_TRACE(sbh, "marking dirty");
4892 ext4_superblock_csum_set(sb);
4893 if (sync)
4894 lock_buffer(sbh);
4895 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
4896 /*
4897 * Oh, dear. A previous attempt to write the
4898 * superblock failed. This could happen because the
4899 * USB device was yanked out. Or it could happen to
4900 * be a transient write error and maybe the block will
4901 * be remapped. Nothing we can do but to retry the
4902 * write and hope for the best.
4903 */
4904 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4905 "superblock detected");
4906 clear_buffer_write_io_error(sbh);
4907 set_buffer_uptodate(sbh);
4908 }
4909 mark_buffer_dirty(sbh);
4910 if (sync) {
4911 unlock_buffer(sbh);
4912 error = __sync_dirty_buffer(sbh,
4913 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
4914 if (error)
4915 return error;
4916
4917 error = buffer_write_io_error(sbh);
4918 if (error) {
4919 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4920 "superblock");
4921 clear_buffer_write_io_error(sbh);
4922 set_buffer_uptodate(sbh);
4923 }
4924 }
4925 return error;
4926 }
4927
4928 /*
4929 * Have we just finished recovery? If so, and if we are mounting (or
4930 * remounting) the filesystem readonly, then we will end up with a
4931 * consistent fs on disk. Record that fact.
4932 */
4933 static void ext4_mark_recovery_complete(struct super_block *sb,
4934 struct ext4_super_block *es)
4935 {
4936 journal_t *journal = EXT4_SB(sb)->s_journal;
4937
4938 if (!ext4_has_feature_journal(sb)) {
4939 BUG_ON(journal != NULL);
4940 return;
4941 }
4942 jbd2_journal_lock_updates(journal);
4943 if (jbd2_journal_flush(journal) < 0)
4944 goto out;
4945
4946 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
4947 ext4_clear_feature_journal_needs_recovery(sb);
4948 ext4_commit_super(sb, 1);
4949 }
4950
4951 out:
4952 jbd2_journal_unlock_updates(journal);
4953 }
4954
4955 /*
4956 * If we are mounting (or read-write remounting) a filesystem whose journal
4957 * has recorded an error from a previous lifetime, move that error to the
4958 * main filesystem now.
4959 */
4960 static void ext4_clear_journal_err(struct super_block *sb,
4961 struct ext4_super_block *es)
4962 {
4963 journal_t *journal;
4964 int j_errno;
4965 const char *errstr;
4966
4967 BUG_ON(!ext4_has_feature_journal(sb));
4968
4969 journal = EXT4_SB(sb)->s_journal;
4970
4971 /*
4972 * Now check for any error status which may have been recorded in the
4973 * journal by a prior ext4_error() or ext4_abort()
4974 */
4975
4976 j_errno = jbd2_journal_errno(journal);
4977 if (j_errno) {
4978 char nbuf[16];
4979
4980 errstr = ext4_decode_error(sb, j_errno, nbuf);
4981 ext4_warning(sb, "Filesystem error recorded "
4982 "from previous mount: %s", errstr);
4983 ext4_warning(sb, "Marking fs in need of filesystem check.");
4984
4985 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4986 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4987 ext4_commit_super(sb, 1);
4988
4989 jbd2_journal_clear_err(journal);
4990 jbd2_journal_update_sb_errno(journal);
4991 }
4992 }
4993
4994 /*
4995 * Force the running and committing transactions to commit,
4996 * and wait on the commit.
4997 */
4998 int ext4_force_commit(struct super_block *sb)
4999 {
5000 journal_t *journal;
5001
5002 if (sb_rdonly(sb))
5003 return 0;
5004
5005 journal = EXT4_SB(sb)->s_journal;
5006 return ext4_journal_force_commit(journal);
5007 }
5008
5009 static int ext4_sync_fs(struct super_block *sb, int wait)
5010 {
5011 int ret = 0;
5012 tid_t target;
5013 bool needs_barrier = false;
5014 struct ext4_sb_info *sbi = EXT4_SB(sb);
5015
5016 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
5017 return 0;
5018
5019 trace_ext4_sync_fs(sb, wait);
5020 flush_workqueue(sbi->rsv_conversion_wq);
5021 /*
5022 * Writeback quota in non-journalled quota case - journalled quota has
5023 * no dirty dquots
5024 */
5025 dquot_writeback_dquots(sb, -1);
5026 /*
5027 * Data writeback is possible w/o journal transaction, so barrier must
5028 * being sent at the end of the function. But we can skip it if
5029 * transaction_commit will do it for us.
5030 */
5031 if (sbi->s_journal) {
5032 target = jbd2_get_latest_transaction(sbi->s_journal);
5033 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5034 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5035 needs_barrier = true;
5036
5037 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5038 if (wait)
5039 ret = jbd2_log_wait_commit(sbi->s_journal,
5040 target);
5041 }
5042 } else if (wait && test_opt(sb, BARRIER))
5043 needs_barrier = true;
5044 if (needs_barrier) {
5045 int err;
5046 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
5047 if (!ret)
5048 ret = err;
5049 }
5050
5051 return ret;
5052 }
5053
5054 /*
5055 * LVM calls this function before a (read-only) snapshot is created. This
5056 * gives us a chance to flush the journal completely and mark the fs clean.
5057 *
5058 * Note that only this function cannot bring a filesystem to be in a clean
5059 * state independently. It relies on upper layer to stop all data & metadata
5060 * modifications.
5061 */
5062 static int ext4_freeze(struct super_block *sb)
5063 {
5064 int error = 0;
5065 journal_t *journal;
5066
5067 if (sb_rdonly(sb))
5068 return 0;
5069
5070 journal = EXT4_SB(sb)->s_journal;
5071
5072 if (journal) {
5073 /* Now we set up the journal barrier. */
5074 jbd2_journal_lock_updates(journal);
5075
5076 /*
5077 * Don't clear the needs_recovery flag if we failed to
5078 * flush the journal.
5079 */
5080 error = jbd2_journal_flush(journal);
5081 if (error < 0)
5082 goto out;
5083
5084 /* Journal blocked and flushed, clear needs_recovery flag. */
5085 ext4_clear_feature_journal_needs_recovery(sb);
5086 }
5087
5088 error = ext4_commit_super(sb, 1);
5089 out:
5090 if (journal)
5091 /* we rely on upper layer to stop further updates */
5092 jbd2_journal_unlock_updates(journal);
5093 return error;
5094 }
5095
5096 /*
5097 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5098 * flag here, even though the filesystem is not technically dirty yet.
5099 */
5100 static int ext4_unfreeze(struct super_block *sb)
5101 {
5102 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5103 return 0;
5104
5105 if (EXT4_SB(sb)->s_journal) {
5106 /* Reset the needs_recovery flag before the fs is unlocked. */
5107 ext4_set_feature_journal_needs_recovery(sb);
5108 }
5109
5110 ext4_commit_super(sb, 1);
5111 return 0;
5112 }
5113
5114 /*
5115 * Structure to save mount options for ext4_remount's benefit
5116 */
5117 struct ext4_mount_options {
5118 unsigned long s_mount_opt;
5119 unsigned long s_mount_opt2;
5120 kuid_t s_resuid;
5121 kgid_t s_resgid;
5122 unsigned long s_commit_interval;
5123 u32 s_min_batch_time, s_max_batch_time;
5124 #ifdef CONFIG_QUOTA
5125 int s_jquota_fmt;
5126 char *s_qf_names[EXT4_MAXQUOTAS];
5127 #endif
5128 };
5129
5130 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5131 {
5132 struct ext4_super_block *es;
5133 struct ext4_sb_info *sbi = EXT4_SB(sb);
5134 unsigned long old_sb_flags;
5135 struct ext4_mount_options old_opts;
5136 int enable_quota = 0;
5137 ext4_group_t g;
5138 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5139 int err = 0;
5140 #ifdef CONFIG_QUOTA
5141 int i, j;
5142 char *to_free[EXT4_MAXQUOTAS];
5143 #endif
5144 char *orig_data = kstrdup(data, GFP_KERNEL);
5145
5146 /* Store the original options */
5147 old_sb_flags = sb->s_flags;
5148 old_opts.s_mount_opt = sbi->s_mount_opt;
5149 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5150 old_opts.s_resuid = sbi->s_resuid;
5151 old_opts.s_resgid = sbi->s_resgid;
5152 old_opts.s_commit_interval = sbi->s_commit_interval;
5153 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5154 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5155 #ifdef CONFIG_QUOTA
5156 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5157 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5158 if (sbi->s_qf_names[i]) {
5159 char *qf_name = get_qf_name(sb, sbi, i);
5160
5161 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5162 if (!old_opts.s_qf_names[i]) {
5163 for (j = 0; j < i; j++)
5164 kfree(old_opts.s_qf_names[j]);
5165 kfree(orig_data);
5166 return -ENOMEM;
5167 }
5168 } else
5169 old_opts.s_qf_names[i] = NULL;
5170 #endif
5171 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5172 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5173
5174 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5175 err = -EINVAL;
5176 goto restore_opts;
5177 }
5178
5179 ext4_clamp_want_extra_isize(sb);
5180
5181 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5182 test_opt(sb, JOURNAL_CHECKSUM)) {
5183 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5184 "during remount not supported; ignoring");
5185 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5186 }
5187
5188 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5189 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5190 ext4_msg(sb, KERN_ERR, "can't mount with "
5191 "both data=journal and delalloc");
5192 err = -EINVAL;
5193 goto restore_opts;
5194 }
5195 if (test_opt(sb, DIOREAD_NOLOCK)) {
5196 ext4_msg(sb, KERN_ERR, "can't mount with "
5197 "both data=journal and dioread_nolock");
5198 err = -EINVAL;
5199 goto restore_opts;
5200 }
5201 if (test_opt(sb, DAX)) {
5202 ext4_msg(sb, KERN_ERR, "can't mount with "
5203 "both data=journal and dax");
5204 err = -EINVAL;
5205 goto restore_opts;
5206 }
5207 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5208 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5209 ext4_msg(sb, KERN_ERR, "can't mount with "
5210 "journal_async_commit in data=ordered mode");
5211 err = -EINVAL;
5212 goto restore_opts;
5213 }
5214 }
5215
5216 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5217 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5218 err = -EINVAL;
5219 goto restore_opts;
5220 }
5221
5222 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5223 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5224 "dax flag with busy inodes while remounting");
5225 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
5226 }
5227
5228 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5229 ext4_abort(sb, "Abort forced by user");
5230
5231 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5232 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5233
5234 es = sbi->s_es;
5235
5236 if (sbi->s_journal) {
5237 ext4_init_journal_params(sb, sbi->s_journal);
5238 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5239 }
5240
5241 if (*flags & SB_LAZYTIME)
5242 sb->s_flags |= SB_LAZYTIME;
5243
5244 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5245 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5246 err = -EROFS;
5247 goto restore_opts;
5248 }
5249
5250 if (*flags & SB_RDONLY) {
5251 err = sync_filesystem(sb);
5252 if (err < 0)
5253 goto restore_opts;
5254 err = dquot_suspend(sb, -1);
5255 if (err < 0)
5256 goto restore_opts;
5257
5258 /*
5259 * First of all, the unconditional stuff we have to do
5260 * to disable replay of the journal when we next remount
5261 */
5262 sb->s_flags |= SB_RDONLY;
5263
5264 /*
5265 * OK, test if we are remounting a valid rw partition
5266 * readonly, and if so set the rdonly flag and then
5267 * mark the partition as valid again.
5268 */
5269 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5270 (sbi->s_mount_state & EXT4_VALID_FS))
5271 es->s_state = cpu_to_le16(sbi->s_mount_state);
5272
5273 if (sbi->s_journal)
5274 ext4_mark_recovery_complete(sb, es);
5275 if (sbi->s_mmp_tsk)
5276 kthread_stop(sbi->s_mmp_tsk);
5277 } else {
5278 /* Make sure we can mount this feature set readwrite */
5279 if (ext4_has_feature_readonly(sb) ||
5280 !ext4_feature_set_ok(sb, 0)) {
5281 err = -EROFS;
5282 goto restore_opts;
5283 }
5284 /*
5285 * Make sure the group descriptor checksums
5286 * are sane. If they aren't, refuse to remount r/w.
5287 */
5288 for (g = 0; g < sbi->s_groups_count; g++) {
5289 struct ext4_group_desc *gdp =
5290 ext4_get_group_desc(sb, g, NULL);
5291
5292 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5293 ext4_msg(sb, KERN_ERR,
5294 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5295 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5296 le16_to_cpu(gdp->bg_checksum));
5297 err = -EFSBADCRC;
5298 goto restore_opts;
5299 }
5300 }
5301
5302 /*
5303 * If we have an unprocessed orphan list hanging
5304 * around from a previously readonly bdev mount,
5305 * require a full umount/remount for now.
5306 */
5307 if (es->s_last_orphan) {
5308 ext4_msg(sb, KERN_WARNING, "Couldn't "
5309 "remount RDWR because of unprocessed "
5310 "orphan inode list. Please "
5311 "umount/remount instead");
5312 err = -EINVAL;
5313 goto restore_opts;
5314 }
5315
5316 /*
5317 * Mounting a RDONLY partition read-write, so reread
5318 * and store the current valid flag. (It may have
5319 * been changed by e2fsck since we originally mounted
5320 * the partition.)
5321 */
5322 if (sbi->s_journal)
5323 ext4_clear_journal_err(sb, es);
5324 sbi->s_mount_state = le16_to_cpu(es->s_state);
5325 if (!ext4_setup_super(sb, es, 0))
5326 sb->s_flags &= ~SB_RDONLY;
5327 if (ext4_has_feature_mmp(sb))
5328 if (ext4_multi_mount_protect(sb,
5329 le64_to_cpu(es->s_mmp_block))) {
5330 err = -EROFS;
5331 goto restore_opts;
5332 }
5333 enable_quota = 1;
5334 }
5335 }
5336
5337 /*
5338 * Reinitialize lazy itable initialization thread based on
5339 * current settings
5340 */
5341 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5342 ext4_unregister_li_request(sb);
5343 else {
5344 ext4_group_t first_not_zeroed;
5345 first_not_zeroed = ext4_has_uninit_itable(sb);
5346 ext4_register_li_request(sb, first_not_zeroed);
5347 }
5348
5349 ext4_setup_system_zone(sb);
5350 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY))
5351 ext4_commit_super(sb, 1);
5352
5353 #ifdef CONFIG_QUOTA
5354 /* Release old quota file names */
5355 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5356 kfree(old_opts.s_qf_names[i]);
5357 if (enable_quota) {
5358 if (sb_any_quota_suspended(sb))
5359 dquot_resume(sb, -1);
5360 else if (ext4_has_feature_quota(sb)) {
5361 err = ext4_enable_quotas(sb);
5362 if (err)
5363 goto restore_opts;
5364 }
5365 }
5366 #endif
5367
5368 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
5369 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5370 kfree(orig_data);
5371 return 0;
5372
5373 restore_opts:
5374 sb->s_flags = old_sb_flags;
5375 sbi->s_mount_opt = old_opts.s_mount_opt;
5376 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5377 sbi->s_resuid = old_opts.s_resuid;
5378 sbi->s_resgid = old_opts.s_resgid;
5379 sbi->s_commit_interval = old_opts.s_commit_interval;
5380 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5381 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5382 #ifdef CONFIG_QUOTA
5383 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5384 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5385 to_free[i] = get_qf_name(sb, sbi, i);
5386 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
5387 }
5388 synchronize_rcu();
5389 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5390 kfree(to_free[i]);
5391 #endif
5392 kfree(orig_data);
5393 return err;
5394 }
5395
5396 #ifdef CONFIG_QUOTA
5397 static int ext4_statfs_project(struct super_block *sb,
5398 kprojid_t projid, struct kstatfs *buf)
5399 {
5400 struct kqid qid;
5401 struct dquot *dquot;
5402 u64 limit;
5403 u64 curblock;
5404
5405 qid = make_kqid_projid(projid);
5406 dquot = dqget(sb, qid);
5407 if (IS_ERR(dquot))
5408 return PTR_ERR(dquot);
5409 spin_lock(&dquot->dq_dqb_lock);
5410
5411 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5412 dquot->dq_dqb.dqb_bsoftlimit :
5413 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5414 if (limit && buf->f_blocks > limit) {
5415 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5416 buf->f_blocks = limit;
5417 buf->f_bfree = buf->f_bavail =
5418 (buf->f_blocks > curblock) ?
5419 (buf->f_blocks - curblock) : 0;
5420 }
5421
5422 limit = dquot->dq_dqb.dqb_isoftlimit ?
5423 dquot->dq_dqb.dqb_isoftlimit :
5424 dquot->dq_dqb.dqb_ihardlimit;
5425 if (limit && buf->f_files > limit) {
5426 buf->f_files = limit;
5427 buf->f_ffree =
5428 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5429 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5430 }
5431
5432 spin_unlock(&dquot->dq_dqb_lock);
5433 dqput(dquot);
5434 return 0;
5435 }
5436 #endif
5437
5438 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5439 {
5440 struct super_block *sb = dentry->d_sb;
5441 struct ext4_sb_info *sbi = EXT4_SB(sb);
5442 struct ext4_super_block *es = sbi->s_es;
5443 ext4_fsblk_t overhead = 0, resv_blocks;
5444 u64 fsid;
5445 s64 bfree;
5446 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5447
5448 if (!test_opt(sb, MINIX_DF))
5449 overhead = sbi->s_overhead;
5450
5451 buf->f_type = EXT4_SUPER_MAGIC;
5452 buf->f_bsize = sb->s_blocksize;
5453 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5454 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5455 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5456 /* prevent underflow in case that few free space is available */
5457 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5458 buf->f_bavail = buf->f_bfree -
5459 (ext4_r_blocks_count(es) + resv_blocks);
5460 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5461 buf->f_bavail = 0;
5462 buf->f_files = le32_to_cpu(es->s_inodes_count);
5463 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5464 buf->f_namelen = EXT4_NAME_LEN;
5465 fsid = le64_to_cpup((void *)es->s_uuid) ^
5466 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5467 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5468 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5469
5470 #ifdef CONFIG_QUOTA
5471 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5472 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5473 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5474 #endif
5475 return 0;
5476 }
5477
5478
5479 #ifdef CONFIG_QUOTA
5480
5481 /*
5482 * Helper functions so that transaction is started before we acquire dqio_sem
5483 * to keep correct lock ordering of transaction > dqio_sem
5484 */
5485 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5486 {
5487 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5488 }
5489
5490 static int ext4_write_dquot(struct dquot *dquot)
5491 {
5492 int ret, err;
5493 handle_t *handle;
5494 struct inode *inode;
5495
5496 inode = dquot_to_inode(dquot);
5497 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5498 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5499 if (IS_ERR(handle))
5500 return PTR_ERR(handle);
5501 ret = dquot_commit(dquot);
5502 err = ext4_journal_stop(handle);
5503 if (!ret)
5504 ret = err;
5505 return ret;
5506 }
5507
5508 static int ext4_acquire_dquot(struct dquot *dquot)
5509 {
5510 int ret, err;
5511 handle_t *handle;
5512
5513 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5514 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5515 if (IS_ERR(handle))
5516 return PTR_ERR(handle);
5517 ret = dquot_acquire(dquot);
5518 err = ext4_journal_stop(handle);
5519 if (!ret)
5520 ret = err;
5521 return ret;
5522 }
5523
5524 static int ext4_release_dquot(struct dquot *dquot)
5525 {
5526 int ret, err;
5527 handle_t *handle;
5528
5529 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5530 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5531 if (IS_ERR(handle)) {
5532 /* Release dquot anyway to avoid endless cycle in dqput() */
5533 dquot_release(dquot);
5534 return PTR_ERR(handle);
5535 }
5536 ret = dquot_release(dquot);
5537 err = ext4_journal_stop(handle);
5538 if (!ret)
5539 ret = err;
5540 return ret;
5541 }
5542
5543 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5544 {
5545 struct super_block *sb = dquot->dq_sb;
5546 struct ext4_sb_info *sbi = EXT4_SB(sb);
5547
5548 /* Are we journaling quotas? */
5549 if (ext4_has_feature_quota(sb) ||
5550 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5551 dquot_mark_dquot_dirty(dquot);
5552 return ext4_write_dquot(dquot);
5553 } else {
5554 return dquot_mark_dquot_dirty(dquot);
5555 }
5556 }
5557
5558 static int ext4_write_info(struct super_block *sb, int type)
5559 {
5560 int ret, err;
5561 handle_t *handle;
5562
5563 /* Data block + inode block */
5564 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5565 if (IS_ERR(handle))
5566 return PTR_ERR(handle);
5567 ret = dquot_commit_info(sb, type);
5568 err = ext4_journal_stop(handle);
5569 if (!ret)
5570 ret = err;
5571 return ret;
5572 }
5573
5574 /*
5575 * Turn on quotas during mount time - we need to find
5576 * the quota file and such...
5577 */
5578 static int ext4_quota_on_mount(struct super_block *sb, int type)
5579 {
5580 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
5581 EXT4_SB(sb)->s_jquota_fmt, type);
5582 }
5583
5584 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5585 {
5586 struct ext4_inode_info *ei = EXT4_I(inode);
5587
5588 /* The first argument of lockdep_set_subclass has to be
5589 * *exactly* the same as the argument to init_rwsem() --- in
5590 * this case, in init_once() --- or lockdep gets unhappy
5591 * because the name of the lock is set using the
5592 * stringification of the argument to init_rwsem().
5593 */
5594 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5595 lockdep_set_subclass(&ei->i_data_sem, subclass);
5596 }
5597
5598 /*
5599 * Standard function to be called on quota_on
5600 */
5601 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5602 const struct path *path)
5603 {
5604 int err;
5605
5606 if (!test_opt(sb, QUOTA))
5607 return -EINVAL;
5608
5609 /* Quotafile not on the same filesystem? */
5610 if (path->dentry->d_sb != sb)
5611 return -EXDEV;
5612 /* Journaling quota? */
5613 if (EXT4_SB(sb)->s_qf_names[type]) {
5614 /* Quotafile not in fs root? */
5615 if (path->dentry->d_parent != sb->s_root)
5616 ext4_msg(sb, KERN_WARNING,
5617 "Quota file not on filesystem root. "
5618 "Journaled quota will not work");
5619 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5620 } else {
5621 /*
5622 * Clear the flag just in case mount options changed since
5623 * last time.
5624 */
5625 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
5626 }
5627
5628 /*
5629 * When we journal data on quota file, we have to flush journal to see
5630 * all updates to the file when we bypass pagecache...
5631 */
5632 if (EXT4_SB(sb)->s_journal &&
5633 ext4_should_journal_data(d_inode(path->dentry))) {
5634 /*
5635 * We don't need to lock updates but journal_flush() could
5636 * otherwise be livelocked...
5637 */
5638 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5639 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5640 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5641 if (err)
5642 return err;
5643 }
5644
5645 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5646 err = dquot_quota_on(sb, type, format_id, path);
5647 if (err) {
5648 lockdep_set_quota_inode(path->dentry->d_inode,
5649 I_DATA_SEM_NORMAL);
5650 } else {
5651 struct inode *inode = d_inode(path->dentry);
5652 handle_t *handle;
5653
5654 /*
5655 * Set inode flags to prevent userspace from messing with quota
5656 * files. If this fails, we return success anyway since quotas
5657 * are already enabled and this is not a hard failure.
5658 */
5659 inode_lock(inode);
5660 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5661 if (IS_ERR(handle))
5662 goto unlock_inode;
5663 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5664 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5665 S_NOATIME | S_IMMUTABLE);
5666 ext4_mark_inode_dirty(handle, inode);
5667 ext4_journal_stop(handle);
5668 unlock_inode:
5669 inode_unlock(inode);
5670 }
5671 return err;
5672 }
5673
5674 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5675 unsigned int flags)
5676 {
5677 int err;
5678 struct inode *qf_inode;
5679 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5680 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5681 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5682 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5683 };
5684
5685 BUG_ON(!ext4_has_feature_quota(sb));
5686
5687 if (!qf_inums[type])
5688 return -EPERM;
5689
5690 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
5691 if (IS_ERR(qf_inode)) {
5692 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5693 return PTR_ERR(qf_inode);
5694 }
5695
5696 /* Don't account quota for quota files to avoid recursion */
5697 qf_inode->i_flags |= S_NOQUOTA;
5698 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5699 err = dquot_enable(qf_inode, type, format_id, flags);
5700 if (err)
5701 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5702 iput(qf_inode);
5703
5704 return err;
5705 }
5706
5707 /* Enable usage tracking for all quota types. */
5708 static int ext4_enable_quotas(struct super_block *sb)
5709 {
5710 int type, err = 0;
5711 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5712 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5713 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5714 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5715 };
5716 bool quota_mopt[EXT4_MAXQUOTAS] = {
5717 test_opt(sb, USRQUOTA),
5718 test_opt(sb, GRPQUOTA),
5719 test_opt(sb, PRJQUOTA),
5720 };
5721
5722 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
5723 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5724 if (qf_inums[type]) {
5725 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5726 DQUOT_USAGE_ENABLED |
5727 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5728 if (err) {
5729 for (type--; type >= 0; type--)
5730 dquot_quota_off(sb, type);
5731
5732 ext4_warning(sb,
5733 "Failed to enable quota tracking "
5734 "(type=%d, err=%d). Please run "
5735 "e2fsck to fix.", type, err);
5736 return err;
5737 }
5738 }
5739 }
5740 return 0;
5741 }
5742
5743 static int ext4_quota_off(struct super_block *sb, int type)
5744 {
5745 struct inode *inode = sb_dqopt(sb)->files[type];
5746 handle_t *handle;
5747 int err;
5748
5749 /* Force all delayed allocation blocks to be allocated.
5750 * Caller already holds s_umount sem */
5751 if (test_opt(sb, DELALLOC))
5752 sync_filesystem(sb);
5753
5754 if (!inode || !igrab(inode))
5755 goto out;
5756
5757 err = dquot_quota_off(sb, type);
5758 if (err || ext4_has_feature_quota(sb))
5759 goto out_put;
5760
5761 inode_lock(inode);
5762 /*
5763 * Update modification times of quota files when userspace can
5764 * start looking at them. If we fail, we return success anyway since
5765 * this is not a hard failure and quotas are already disabled.
5766 */
5767 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5768 if (IS_ERR(handle))
5769 goto out_unlock;
5770 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5771 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
5772 inode->i_mtime = inode->i_ctime = current_time(inode);
5773 ext4_mark_inode_dirty(handle, inode);
5774 ext4_journal_stop(handle);
5775 out_unlock:
5776 inode_unlock(inode);
5777 out_put:
5778 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
5779 iput(inode);
5780 return err;
5781 out:
5782 return dquot_quota_off(sb, type);
5783 }
5784
5785 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5786 * acquiring the locks... As quota files are never truncated and quota code
5787 * itself serializes the operations (and no one else should touch the files)
5788 * we don't have to be afraid of races */
5789 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5790 size_t len, loff_t off)
5791 {
5792 struct inode *inode = sb_dqopt(sb)->files[type];
5793 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5794 int offset = off & (sb->s_blocksize - 1);
5795 int tocopy;
5796 size_t toread;
5797 struct buffer_head *bh;
5798 loff_t i_size = i_size_read(inode);
5799
5800 if (off > i_size)
5801 return 0;
5802 if (off+len > i_size)
5803 len = i_size-off;
5804 toread = len;
5805 while (toread > 0) {
5806 tocopy = sb->s_blocksize - offset < toread ?
5807 sb->s_blocksize - offset : toread;
5808 bh = ext4_bread(NULL, inode, blk, 0);
5809 if (IS_ERR(bh))
5810 return PTR_ERR(bh);
5811 if (!bh) /* A hole? */
5812 memset(data, 0, tocopy);
5813 else
5814 memcpy(data, bh->b_data+offset, tocopy);
5815 brelse(bh);
5816 offset = 0;
5817 toread -= tocopy;
5818 data += tocopy;
5819 blk++;
5820 }
5821 return len;
5822 }
5823
5824 /* Write to quotafile (we know the transaction is already started and has
5825 * enough credits) */
5826 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5827 const char *data, size_t len, loff_t off)
5828 {
5829 struct inode *inode = sb_dqopt(sb)->files[type];
5830 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5831 int err, offset = off & (sb->s_blocksize - 1);
5832 int retries = 0;
5833 struct buffer_head *bh;
5834 handle_t *handle = journal_current_handle();
5835
5836 if (EXT4_SB(sb)->s_journal && !handle) {
5837 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5838 " cancelled because transaction is not started",
5839 (unsigned long long)off, (unsigned long long)len);
5840 return -EIO;
5841 }
5842 /*
5843 * Since we account only one data block in transaction credits,
5844 * then it is impossible to cross a block boundary.
5845 */
5846 if (sb->s_blocksize - offset < len) {
5847 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5848 " cancelled because not block aligned",
5849 (unsigned long long)off, (unsigned long long)len);
5850 return -EIO;
5851 }
5852
5853 do {
5854 bh = ext4_bread(handle, inode, blk,
5855 EXT4_GET_BLOCKS_CREATE |
5856 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5857 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5858 ext4_should_retry_alloc(inode->i_sb, &retries));
5859 if (IS_ERR(bh))
5860 return PTR_ERR(bh);
5861 if (!bh)
5862 goto out;
5863 BUFFER_TRACE(bh, "get write access");
5864 err = ext4_journal_get_write_access(handle, bh);
5865 if (err) {
5866 brelse(bh);
5867 return err;
5868 }
5869 lock_buffer(bh);
5870 memcpy(bh->b_data+offset, data, len);
5871 flush_dcache_page(bh->b_page);
5872 unlock_buffer(bh);
5873 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5874 brelse(bh);
5875 out:
5876 if (inode->i_size < off + len) {
5877 i_size_write(inode, off + len);
5878 EXT4_I(inode)->i_disksize = inode->i_size;
5879 ext4_mark_inode_dirty(handle, inode);
5880 }
5881 return len;
5882 }
5883
5884 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5885 {
5886 const struct quota_format_ops *ops;
5887
5888 if (!sb_has_quota_loaded(sb, qid->type))
5889 return -ESRCH;
5890 ops = sb_dqopt(sb)->ops[qid->type];
5891 if (!ops || !ops->get_next_id)
5892 return -ENOSYS;
5893 return dquot_get_next_id(sb, qid);
5894 }
5895 #endif
5896
5897 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5898 const char *dev_name, void *data)
5899 {
5900 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5901 }
5902
5903 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5904 static inline void register_as_ext2(void)
5905 {
5906 int err = register_filesystem(&ext2_fs_type);
5907 if (err)
5908 printk(KERN_WARNING
5909 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5910 }
5911
5912 static inline void unregister_as_ext2(void)
5913 {
5914 unregister_filesystem(&ext2_fs_type);
5915 }
5916
5917 static inline int ext2_feature_set_ok(struct super_block *sb)
5918 {
5919 if (ext4_has_unknown_ext2_incompat_features(sb))
5920 return 0;
5921 if (sb_rdonly(sb))
5922 return 1;
5923 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5924 return 0;
5925 return 1;
5926 }
5927 #else
5928 static inline void register_as_ext2(void) { }
5929 static inline void unregister_as_ext2(void) { }
5930 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5931 #endif
5932
5933 static inline void register_as_ext3(void)
5934 {
5935 int err = register_filesystem(&ext3_fs_type);
5936 if (err)
5937 printk(KERN_WARNING
5938 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5939 }
5940
5941 static inline void unregister_as_ext3(void)
5942 {
5943 unregister_filesystem(&ext3_fs_type);
5944 }
5945
5946 static inline int ext3_feature_set_ok(struct super_block *sb)
5947 {
5948 if (ext4_has_unknown_ext3_incompat_features(sb))
5949 return 0;
5950 if (!ext4_has_feature_journal(sb))
5951 return 0;
5952 if (sb_rdonly(sb))
5953 return 1;
5954 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5955 return 0;
5956 return 1;
5957 }
5958
5959 static struct file_system_type ext4_fs_type = {
5960 .owner = THIS_MODULE,
5961 .name = "ext4",
5962 .mount = ext4_mount,
5963 .kill_sb = kill_block_super,
5964 .fs_flags = FS_REQUIRES_DEV | FS_USERNS_MOUNT,
5965 };
5966 MODULE_ALIAS_FS("ext4");
5967
5968 /* Shared across all ext4 file systems */
5969 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5970
5971 static int __init ext4_init_fs(void)
5972 {
5973 int i, err;
5974
5975 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5976 ext4_li_info = NULL;
5977 mutex_init(&ext4_li_mtx);
5978
5979 /* Build-time check for flags consistency */
5980 ext4_check_flag_values();
5981
5982 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5983 init_waitqueue_head(&ext4__ioend_wq[i]);
5984
5985 err = ext4_init_es();
5986 if (err)
5987 return err;
5988
5989 err = ext4_init_pageio();
5990 if (err)
5991 goto out5;
5992
5993 err = ext4_init_system_zone();
5994 if (err)
5995 goto out4;
5996
5997 err = ext4_init_sysfs();
5998 if (err)
5999 goto out3;
6000
6001 err = ext4_init_mballoc();
6002 if (err)
6003 goto out2;
6004 err = init_inodecache();
6005 if (err)
6006 goto out1;
6007 register_as_ext3();
6008 register_as_ext2();
6009 err = register_filesystem(&ext4_fs_type);
6010 if (err)
6011 goto out;
6012
6013 return 0;
6014 out:
6015 unregister_as_ext2();
6016 unregister_as_ext3();
6017 destroy_inodecache();
6018 out1:
6019 ext4_exit_mballoc();
6020 out2:
6021 ext4_exit_sysfs();
6022 out3:
6023 ext4_exit_system_zone();
6024 out4:
6025 ext4_exit_pageio();
6026 out5:
6027 ext4_exit_es();
6028
6029 return err;
6030 }
6031
6032 static void __exit ext4_exit_fs(void)
6033 {
6034 ext4_destroy_lazyinit_thread();
6035 unregister_as_ext2();
6036 unregister_as_ext3();
6037 unregister_filesystem(&ext4_fs_type);
6038 destroy_inodecache();
6039 ext4_exit_mballoc();
6040 ext4_exit_sysfs();
6041 ext4_exit_system_zone();
6042 ext4_exit_pageio();
6043 ext4_exit_es();
6044 }
6045
6046 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6047 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6048 MODULE_LICENSE("GPL");
6049 MODULE_SOFTDEP("pre: crc32c");
6050 module_init(ext4_init_fs)
6051 module_exit(ext4_exit_fs)
ubuntu-bionic-kernel mirror