1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *  linux/fs/ext4/super.c
4  *
5  * Copyright (C) 1992, 1993, 1994, 1995
6  * Remy Card (card@masi.ibp.fr)
7  * Laboratoire MASI - Institut Blaise Pascal
8  * Universite Pierre et Marie Curie (Paris VI)
9  *
10  *  from
11  *
12  *  linux/fs/minix/inode.c
13  *
14  *  Copyright (C) 1991, 1992  Linus Torvalds
15  *
16  *  Big-endian to little-endian byte-swapping/bitmaps by
17  *        David S. Miller (davem@caip.rutgers.edu), 1995
18  */
19 
20 #include <linux/module.h>
21 #include <linux/string.h>
22 #include <linux/fs.h>
23 #include <linux/time.h>
24 #include <linux/vmalloc.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/backing-dev.h>
29 #include <linux/parser.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/dax.h>
42 #include <linux/cleancache.h>
43 #include <linux/uaccess.h>
44 #include <linux/iversion.h>
45 
46 #include <linux/kthread.h>
47 #include <linux/freezer.h>
48 
49 #include "ext4.h"
50 #include "ext4_extents.h"	/* Needed for trace points definition */
51 #include "ext4_jbd2.h"
52 #include "xattr.h"
53 #include "acl.h"
54 #include "mballoc.h"
55 #include "fsmap.h"
56 
57 #define CREATE_TRACE_POINTS
58 #include <trace/events/ext4.h>
59 
60 static struct ext4_lazy_init *ext4_li_info;
61 static struct mutex ext4_li_mtx;
62 static struct ratelimit_state ext4_mount_msg_ratelimit;
63 
64 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
65 			     unsigned long journal_devnum);
66 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
67 static int ext4_commit_super(struct super_block *sb, int sync);
68 static int ext4_mark_recovery_complete(struct super_block *sb,
69 					struct ext4_super_block *es);
70 static int ext4_clear_journal_err(struct super_block *sb,
71 				  struct ext4_super_block *es);
72 static int ext4_sync_fs(struct super_block *sb, int wait);
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 		       const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85 static struct inode *ext4_get_journal_inode(struct super_block *sb,
86 					    unsigned int journal_inum);
87 
88 /*
89  * Lock ordering
90  *
91  * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
92  * i_mmap_rwsem (inode->i_mmap_rwsem)!
93  *
94  * page fault path:
95  * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
96  *   page lock -> i_data_sem (rw)
97  *
98  * buffered write path:
99  * sb_start_write -> i_mutex -> mmap_sem
100  * sb_start_write -> i_mutex -> transaction start -> page lock ->
101  *   i_data_sem (rw)
102  *
103  * truncate:
104  * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
105  * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
106  *   i_data_sem (rw)
107  *
108  * direct IO:
109  * sb_start_write -> i_mutex -> mmap_sem
110  * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
111  *
112  * writepages:
113  * transaction start -> page lock(s) -> i_data_sem (rw)
114  */
115 
116 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
117 static struct file_system_type ext2_fs_type = {
118 	.owner		= THIS_MODULE,
119 	.name		= "ext2",
120 	.mount		= ext4_mount,
121 	.kill_sb	= kill_block_super,
122 	.fs_flags	= FS_REQUIRES_DEV,
123 };
124 MODULE_ALIAS_FS("ext2");
125 MODULE_ALIAS("ext2");
126 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
127 #else
128 #define IS_EXT2_SB(sb) (0)
129 #endif
130 
131 
132 static struct file_system_type ext3_fs_type = {
133 	.owner		= THIS_MODULE,
134 	.name		= "ext3",
135 	.mount		= ext4_mount,
136 	.kill_sb	= kill_block_super,
137 	.fs_flags	= FS_REQUIRES_DEV,
138 };
139 MODULE_ALIAS_FS("ext3");
140 MODULE_ALIAS("ext3");
141 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
142 
143 /*
144  * This works like sb_bread() except it uses ERR_PTR for error
145  * returns.  Currently with sb_bread it's impossible to distinguish
146  * between ENOMEM and EIO situations (since both result in a NULL
147  * return.
148  */
149 struct buffer_head *
ext4_sb_bread(struct super_block * sb,sector_t block,int op_flags)150 ext4_sb_bread(struct super_block *sb, sector_t block, int op_flags)
151 {
152 	struct buffer_head *bh = sb_getblk(sb, block);
153 
154 	if (bh == NULL)
155 		return ERR_PTR(-ENOMEM);
156 	if (buffer_uptodate(bh))
157 		return bh;
158 	ll_rw_block(REQ_OP_READ, REQ_META | op_flags, 1, &bh);
159 	wait_on_buffer(bh);
160 	if (buffer_uptodate(bh))
161 		return bh;
162 	put_bh(bh);
163 	return ERR_PTR(-EIO);
164 }
165 
ext4_verify_csum_type(struct super_block * sb,struct ext4_super_block * es)166 static int ext4_verify_csum_type(struct super_block *sb,
167 				 struct ext4_super_block *es)
168 {
169 	if (!ext4_has_feature_metadata_csum(sb))
170 		return 1;
171 
172 	return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
173 }
174 
ext4_superblock_csum(struct super_block * sb,struct ext4_super_block * es)175 static __le32 ext4_superblock_csum(struct super_block *sb,
176 				   struct ext4_super_block *es)
177 {
178 	struct ext4_sb_info *sbi = EXT4_SB(sb);
179 	int offset = offsetof(struct ext4_super_block, s_checksum);
180 	__u32 csum;
181 
182 	csum = ext4_chksum(sbi, ~0, (char *)es, offset);
183 
184 	return cpu_to_le32(csum);
185 }
186 
ext4_superblock_csum_verify(struct super_block * sb,struct ext4_super_block * es)187 static int ext4_superblock_csum_verify(struct super_block *sb,
188 				       struct ext4_super_block *es)
189 {
190 	if (!ext4_has_metadata_csum(sb))
191 		return 1;
192 
193 	return es->s_checksum == ext4_superblock_csum(sb, es);
194 }
195 
ext4_superblock_csum_set(struct super_block * sb)196 void ext4_superblock_csum_set(struct super_block *sb)
197 {
198 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
199 
200 	if (!ext4_has_metadata_csum(sb))
201 		return;
202 
203 	es->s_checksum = ext4_superblock_csum(sb, es);
204 }
205 
ext4_kvmalloc(size_t size,gfp_t flags)206 void *ext4_kvmalloc(size_t size, gfp_t flags)
207 {
208 	void *ret;
209 
210 	ret = kmalloc(size, flags | __GFP_NOWARN);
211 	if (!ret)
212 		ret = __vmalloc(size, flags, PAGE_KERNEL);
213 	return ret;
214 }
215 
ext4_kvzalloc(size_t size,gfp_t flags)216 void *ext4_kvzalloc(size_t size, gfp_t flags)
217 {
218 	void *ret;
219 
220 	ret = kzalloc(size, flags | __GFP_NOWARN);
221 	if (!ret)
222 		ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
223 	return ret;
224 }
225 
ext4_block_bitmap(struct super_block * sb,struct ext4_group_desc * bg)226 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
227 			       struct ext4_group_desc *bg)
228 {
229 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
230 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
231 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
232 }
233 
ext4_inode_bitmap(struct super_block * sb,struct ext4_group_desc * bg)234 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
235 			       struct ext4_group_desc *bg)
236 {
237 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
238 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
239 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
240 }
241 
ext4_inode_table(struct super_block * sb,struct ext4_group_desc * bg)242 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
243 			      struct ext4_group_desc *bg)
244 {
245 	return le32_to_cpu(bg->bg_inode_table_lo) |
246 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
247 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
248 }
249 
ext4_free_group_clusters(struct super_block * sb,struct ext4_group_desc * bg)250 __u32 ext4_free_group_clusters(struct super_block *sb,
251 			       struct ext4_group_desc *bg)
252 {
253 	return le16_to_cpu(bg->bg_free_blocks_count_lo) |
254 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
255 		 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
256 }
257 
ext4_free_inodes_count(struct super_block * sb,struct ext4_group_desc * bg)258 __u32 ext4_free_inodes_count(struct super_block *sb,
259 			      struct ext4_group_desc *bg)
260 {
261 	return le16_to_cpu(bg->bg_free_inodes_count_lo) |
262 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
263 		 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
264 }
265 
ext4_used_dirs_count(struct super_block * sb,struct ext4_group_desc * bg)266 __u32 ext4_used_dirs_count(struct super_block *sb,
267 			      struct ext4_group_desc *bg)
268 {
269 	return le16_to_cpu(bg->bg_used_dirs_count_lo) |
270 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
271 		 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
272 }
273 
ext4_itable_unused_count(struct super_block * sb,struct ext4_group_desc * bg)274 __u32 ext4_itable_unused_count(struct super_block *sb,
275 			      struct ext4_group_desc *bg)
276 {
277 	return le16_to_cpu(bg->bg_itable_unused_lo) |
278 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
279 		 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
280 }
281 
ext4_block_bitmap_set(struct super_block * sb,struct ext4_group_desc * bg,ext4_fsblk_t blk)282 void ext4_block_bitmap_set(struct super_block *sb,
283 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
284 {
285 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
286 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
288 }
289 
ext4_inode_bitmap_set(struct super_block * sb,struct ext4_group_desc * bg,ext4_fsblk_t blk)290 void ext4_inode_bitmap_set(struct super_block *sb,
291 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
292 {
293 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
294 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
295 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
296 }
297 
ext4_inode_table_set(struct super_block * sb,struct ext4_group_desc * bg,ext4_fsblk_t blk)298 void ext4_inode_table_set(struct super_block *sb,
299 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
300 {
301 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
302 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
303 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
304 }
305 
ext4_free_group_clusters_set(struct super_block * sb,struct ext4_group_desc * bg,__u32 count)306 void ext4_free_group_clusters_set(struct super_block *sb,
307 				  struct ext4_group_desc *bg, __u32 count)
308 {
309 	bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
310 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
311 		bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
312 }
313 
ext4_free_inodes_set(struct super_block * sb,struct ext4_group_desc * bg,__u32 count)314 void ext4_free_inodes_set(struct super_block *sb,
315 			  struct ext4_group_desc *bg, __u32 count)
316 {
317 	bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
318 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
319 		bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
320 }
321 
ext4_used_dirs_set(struct super_block * sb,struct ext4_group_desc * bg,__u32 count)322 void ext4_used_dirs_set(struct super_block *sb,
323 			  struct ext4_group_desc *bg, __u32 count)
324 {
325 	bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
326 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
327 		bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
328 }
329 
ext4_itable_unused_set(struct super_block * sb,struct ext4_group_desc * bg,__u32 count)330 void ext4_itable_unused_set(struct super_block *sb,
331 			  struct ext4_group_desc *bg, __u32 count)
332 {
333 	bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
334 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
335 		bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
336 }
337 
__ext4_update_tstamp(__le32 * lo,__u8 * hi)338 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi)
339 {
340 	time64_t now = ktime_get_real_seconds();
341 
342 	now = clamp_val(now, 0, (1ull << 40) - 1);
343 
344 	*lo = cpu_to_le32(lower_32_bits(now));
345 	*hi = upper_32_bits(now);
346 }
347 
__ext4_get_tstamp(__le32 * lo,__u8 * hi)348 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
349 {
350 	return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
351 }
352 #define ext4_update_tstamp(es, tstamp) \
353 	__ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
354 #define ext4_get_tstamp(es, tstamp) \
355 	__ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
356 
__save_error_info(struct super_block * sb,const char * func,unsigned int line)357 static void __save_error_info(struct super_block *sb, const char *func,
358 			    unsigned int line)
359 {
360 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
361 
362 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
363 	if (bdev_read_only(sb->s_bdev))
364 		return;
365 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
366 	ext4_update_tstamp(es, s_last_error_time);
367 	strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
368 	es->s_last_error_line = cpu_to_le32(line);
369 	if (!es->s_first_error_time) {
370 		es->s_first_error_time = es->s_last_error_time;
371 		es->s_first_error_time_hi = es->s_last_error_time_hi;
372 		strncpy(es->s_first_error_func, func,
373 			sizeof(es->s_first_error_func));
374 		es->s_first_error_line = cpu_to_le32(line);
375 		es->s_first_error_ino = es->s_last_error_ino;
376 		es->s_first_error_block = es->s_last_error_block;
377 	}
378 	/*
379 	 * Start the daily error reporting function if it hasn't been
380 	 * started already
381 	 */
382 	if (!es->s_error_count)
383 		mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
384 	le32_add_cpu(&es->s_error_count, 1);
385 }
386 
save_error_info(struct super_block * sb,const char * func,unsigned int line)387 static void save_error_info(struct super_block *sb, const char *func,
388 			    unsigned int line)
389 {
390 	__save_error_info(sb, func, line);
391 	if (!bdev_read_only(sb->s_bdev))
392 		ext4_commit_super(sb, 1);
393 }
394 
395 /*
396  * The del_gendisk() function uninitializes the disk-specific data
397  * structures, including the bdi structure, without telling anyone
398  * else.  Once this happens, any attempt to call mark_buffer_dirty()
399  * (for example, by ext4_commit_super), will cause a kernel OOPS.
400  * This is a kludge to prevent these oops until we can put in a proper
401  * hook in del_gendisk() to inform the VFS and file system layers.
402  */
block_device_ejected(struct super_block * sb)403 static int block_device_ejected(struct super_block *sb)
404 {
405 	struct inode *bd_inode = sb->s_bdev->bd_inode;
406 	struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
407 
408 	return bdi->dev == NULL;
409 }
410 
ext4_journal_commit_callback(journal_t * journal,transaction_t * txn)411 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
412 {
413 	struct super_block		*sb = journal->j_private;
414 	struct ext4_sb_info		*sbi = EXT4_SB(sb);
415 	int				error = is_journal_aborted(journal);
416 	struct ext4_journal_cb_entry	*jce;
417 
418 	BUG_ON(txn->t_state == T_FINISHED);
419 
420 	ext4_process_freed_data(sb, txn->t_tid);
421 
422 	spin_lock(&sbi->s_md_lock);
423 	while (!list_empty(&txn->t_private_list)) {
424 		jce = list_entry(txn->t_private_list.next,
425 				 struct ext4_journal_cb_entry, jce_list);
426 		list_del_init(&jce->jce_list);
427 		spin_unlock(&sbi->s_md_lock);
428 		jce->jce_func(sb, jce, error);
429 		spin_lock(&sbi->s_md_lock);
430 	}
431 	spin_unlock(&sbi->s_md_lock);
432 }
433 
system_going_down(void)434 static bool system_going_down(void)
435 {
436 	return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
437 		|| system_state == SYSTEM_RESTART;
438 }
439 
440 /* Deal with the reporting of failure conditions on a filesystem such as
441  * inconsistencies detected or read IO failures.
442  *
443  * On ext2, we can store the error state of the filesystem in the
444  * superblock.  That is not possible on ext4, because we may have other
445  * write ordering constraints on the superblock which prevent us from
446  * writing it out straight away; and given that the journal is about to
447  * be aborted, we can't rely on the current, or future, transactions to
448  * write out the superblock safely.
449  *
450  * We'll just use the jbd2_journal_abort() error code to record an error in
451  * the journal instead.  On recovery, the journal will complain about
452  * that error until we've noted it down and cleared it.
453  */
454 
ext4_handle_error(struct super_block * sb)455 static void ext4_handle_error(struct super_block *sb)
456 {
457 	journal_t *journal = EXT4_SB(sb)->s_journal;
458 
459 	if (test_opt(sb, WARN_ON_ERROR))
460 		WARN_ON_ONCE(1);
461 
462 	if (sb_rdonly(sb) || test_opt(sb, ERRORS_CONT))
463 		return;
464 
465 	EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
466 	if (journal)
467 		jbd2_journal_abort(journal, -EIO);
468 	/*
469 	 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
470 	 * could panic during 'reboot -f' as the underlying device got already
471 	 * disabled.
472 	 */
473 	if (test_opt(sb, ERRORS_RO) || system_going_down()) {
474 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
475 		/*
476 		 * Make sure updated value of ->s_mount_flags will be visible
477 		 * before ->s_flags update
478 		 */
479 		smp_wmb();
480 		sb->s_flags |= SB_RDONLY;
481 	} else if (test_opt(sb, ERRORS_PANIC)) {
482 		if (EXT4_SB(sb)->s_journal &&
483 		  !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
484 			return;
485 		panic("EXT4-fs (device %s): panic forced after error\n",
486 			sb->s_id);
487 	}
488 }
489 
490 #define ext4_error_ratelimit(sb)					\
491 		___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state),	\
492 			     "EXT4-fs error")
493 
__ext4_error(struct super_block * sb,const char * function,unsigned int line,const char * fmt,...)494 void __ext4_error(struct super_block *sb, const char *function,
495 		  unsigned int line, const char *fmt, ...)
496 {
497 	struct va_format vaf;
498 	va_list args;
499 
500 	if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
501 		return;
502 
503 	trace_ext4_error(sb, function, line);
504 	if (ext4_error_ratelimit(sb)) {
505 		va_start(args, fmt);
506 		vaf.fmt = fmt;
507 		vaf.va = &args;
508 		printk(KERN_CRIT
509 		       "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
510 		       sb->s_id, function, line, current->comm, &vaf);
511 		va_end(args);
512 	}
513 	save_error_info(sb, function, line);
514 	ext4_handle_error(sb);
515 }
516 
__ext4_error_inode(struct inode * inode,const char * function,unsigned int line,ext4_fsblk_t block,const char * fmt,...)517 void __ext4_error_inode(struct inode *inode, const char *function,
518 			unsigned int line, ext4_fsblk_t block,
519 			const char *fmt, ...)
520 {
521 	va_list args;
522 	struct va_format vaf;
523 	struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
524 
525 	if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
526 		return;
527 
528 	trace_ext4_error(inode->i_sb, function, line);
529 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
530 	es->s_last_error_block = cpu_to_le64(block);
531 	if (ext4_error_ratelimit(inode->i_sb)) {
532 		va_start(args, fmt);
533 		vaf.fmt = fmt;
534 		vaf.va = &args;
535 		if (block)
536 			printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
537 			       "inode #%lu: block %llu: comm %s: %pV\n",
538 			       inode->i_sb->s_id, function, line, inode->i_ino,
539 			       block, current->comm, &vaf);
540 		else
541 			printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
542 			       "inode #%lu: comm %s: %pV\n",
543 			       inode->i_sb->s_id, function, line, inode->i_ino,
544 			       current->comm, &vaf);
545 		va_end(args);
546 	}
547 	save_error_info(inode->i_sb, function, line);
548 	ext4_handle_error(inode->i_sb);
549 }
550 
__ext4_error_file(struct file * file,const char * function,unsigned int line,ext4_fsblk_t block,const char * fmt,...)551 void __ext4_error_file(struct file *file, const char *function,
552 		       unsigned int line, ext4_fsblk_t block,
553 		       const char *fmt, ...)
554 {
555 	va_list args;
556 	struct va_format vaf;
557 	struct ext4_super_block *es;
558 	struct inode *inode = file_inode(file);
559 	char pathname[80], *path;
560 
561 	if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
562 		return;
563 
564 	trace_ext4_error(inode->i_sb, function, line);
565 	es = EXT4_SB(inode->i_sb)->s_es;
566 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
567 	if (ext4_error_ratelimit(inode->i_sb)) {
568 		path = file_path(file, pathname, sizeof(pathname));
569 		if (IS_ERR(path))
570 			path = "(unknown)";
571 		va_start(args, fmt);
572 		vaf.fmt = fmt;
573 		vaf.va = &args;
574 		if (block)
575 			printk(KERN_CRIT
576 			       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
577 			       "block %llu: comm %s: path %s: %pV\n",
578 			       inode->i_sb->s_id, function, line, inode->i_ino,
579 			       block, current->comm, path, &vaf);
580 		else
581 			printk(KERN_CRIT
582 			       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
583 			       "comm %s: path %s: %pV\n",
584 			       inode->i_sb->s_id, function, line, inode->i_ino,
585 			       current->comm, path, &vaf);
586 		va_end(args);
587 	}
588 	save_error_info(inode->i_sb, function, line);
589 	ext4_handle_error(inode->i_sb);
590 }
591 
ext4_decode_error(struct super_block * sb,int errno,char nbuf[16])592 const char *ext4_decode_error(struct super_block *sb, int errno,
593 			      char nbuf[16])
594 {
595 	char *errstr = NULL;
596 
597 	switch (errno) {
598 	case -EFSCORRUPTED:
599 		errstr = "Corrupt filesystem";
600 		break;
601 	case -EFSBADCRC:
602 		errstr = "Filesystem failed CRC";
603 		break;
604 	case -EIO:
605 		errstr = "IO failure";
606 		break;
607 	case -ENOMEM:
608 		errstr = "Out of memory";
609 		break;
610 	case -EROFS:
611 		if (!sb || (EXT4_SB(sb)->s_journal &&
612 			    EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
613 			errstr = "Journal has aborted";
614 		else
615 			errstr = "Readonly filesystem";
616 		break;
617 	default:
618 		/* If the caller passed in an extra buffer for unknown
619 		 * errors, textualise them now.  Else we just return
620 		 * NULL. */
621 		if (nbuf) {
622 			/* Check for truncated error codes... */
623 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
624 				errstr = nbuf;
625 		}
626 		break;
627 	}
628 
629 	return errstr;
630 }
631 
632 /* __ext4_std_error decodes expected errors from journaling functions
633  * automatically and invokes the appropriate error response.  */
634 
__ext4_std_error(struct super_block * sb,const char * function,unsigned int line,int errno)635 void __ext4_std_error(struct super_block *sb, const char *function,
636 		      unsigned int line, int errno)
637 {
638 	char nbuf[16];
639 	const char *errstr;
640 
641 	if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
642 		return;
643 
644 	/* Special case: if the error is EROFS, and we're not already
645 	 * inside a transaction, then there's really no point in logging
646 	 * an error. */
647 	if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
648 		return;
649 
650 	if (ext4_error_ratelimit(sb)) {
651 		errstr = ext4_decode_error(sb, errno, nbuf);
652 		printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
653 		       sb->s_id, function, line, errstr);
654 	}
655 
656 	save_error_info(sb, function, line);
657 	ext4_handle_error(sb);
658 }
659 
660 /*
661  * ext4_abort is a much stronger failure handler than ext4_error.  The
662  * abort function may be used to deal with unrecoverable failures such
663  * as journal IO errors or ENOMEM at a critical moment in log management.
664  *
665  * We unconditionally force the filesystem into an ABORT|READONLY state,
666  * unless the error response on the fs has been set to panic in which
667  * case we take the easy way out and panic immediately.
668  */
669 
__ext4_abort(struct super_block * sb,const char * function,unsigned int line,const char * fmt,...)670 void __ext4_abort(struct super_block *sb, const char *function,
671 		unsigned int line, const char *fmt, ...)
672 {
673 	struct va_format vaf;
674 	va_list args;
675 
676 	if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
677 		return;
678 
679 	save_error_info(sb, function, line);
680 	va_start(args, fmt);
681 	vaf.fmt = fmt;
682 	vaf.va = &args;
683 	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
684 	       sb->s_id, function, line, &vaf);
685 	va_end(args);
686 
687 	if (sb_rdonly(sb) == 0) {
688 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
689 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
690 		/*
691 		 * Make sure updated value of ->s_mount_flags will be visible
692 		 * before ->s_flags update
693 		 */
694 		smp_wmb();
695 		sb->s_flags |= SB_RDONLY;
696 		if (EXT4_SB(sb)->s_journal)
697 			jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
698 		save_error_info(sb, function, line);
699 	}
700 	if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
701 		if (EXT4_SB(sb)->s_journal &&
702 		  !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
703 			return;
704 		panic("EXT4-fs panic from previous error\n");
705 	}
706 }
707 
__ext4_msg(struct super_block * sb,const char * prefix,const char * fmt,...)708 void __ext4_msg(struct super_block *sb,
709 		const char *prefix, const char *fmt, ...)
710 {
711 	struct va_format vaf;
712 	va_list args;
713 
714 	if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
715 		return;
716 
717 	va_start(args, fmt);
718 	vaf.fmt = fmt;
719 	vaf.va = &args;
720 	printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
721 	va_end(args);
722 }
723 
724 #define ext4_warning_ratelimit(sb)					\
725 		___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),	\
726 			     "EXT4-fs warning")
727 
__ext4_warning(struct super_block * sb,const char * function,unsigned int line,const char * fmt,...)728 void __ext4_warning(struct super_block *sb, const char *function,
729 		    unsigned int line, const char *fmt, ...)
730 {
731 	struct va_format vaf;
732 	va_list args;
733 
734 	if (!ext4_warning_ratelimit(sb))
735 		return;
736 
737 	va_start(args, fmt);
738 	vaf.fmt = fmt;
739 	vaf.va = &args;
740 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
741 	       sb->s_id, function, line, &vaf);
742 	va_end(args);
743 }
744 
__ext4_warning_inode(const struct inode * inode,const char * function,unsigned int line,const char * fmt,...)745 void __ext4_warning_inode(const struct inode *inode, const char *function,
746 			  unsigned int line, const char *fmt, ...)
747 {
748 	struct va_format vaf;
749 	va_list args;
750 
751 	if (!ext4_warning_ratelimit(inode->i_sb))
752 		return;
753 
754 	va_start(args, fmt);
755 	vaf.fmt = fmt;
756 	vaf.va = &args;
757 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
758 	       "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
759 	       function, line, inode->i_ino, current->comm, &vaf);
760 	va_end(args);
761 }
762 
__ext4_grp_locked_error(const char * function,unsigned int line,struct super_block * sb,ext4_group_t grp,unsigned long ino,ext4_fsblk_t block,const char * fmt,...)763 void __ext4_grp_locked_error(const char *function, unsigned int line,
764 			     struct super_block *sb, ext4_group_t grp,
765 			     unsigned long ino, ext4_fsblk_t block,
766 			     const char *fmt, ...)
767 __releases(bitlock)
768 __acquires(bitlock)
769 {
770 	struct va_format vaf;
771 	va_list args;
772 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
773 
774 	if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
775 		return;
776 
777 	trace_ext4_error(sb, function, line);
778 	es->s_last_error_ino = cpu_to_le32(ino);
779 	es->s_last_error_block = cpu_to_le64(block);
780 	__save_error_info(sb, function, line);
781 
782 	if (ext4_error_ratelimit(sb)) {
783 		va_start(args, fmt);
784 		vaf.fmt = fmt;
785 		vaf.va = &args;
786 		printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
787 		       sb->s_id, function, line, grp);
788 		if (ino)
789 			printk(KERN_CONT "inode %lu: ", ino);
790 		if (block)
791 			printk(KERN_CONT "block %llu:",
792 			       (unsigned long long) block);
793 		printk(KERN_CONT "%pV\n", &vaf);
794 		va_end(args);
795 	}
796 
797 	if (test_opt(sb, WARN_ON_ERROR))
798 		WARN_ON_ONCE(1);
799 
800 	if (test_opt(sb, ERRORS_CONT)) {
801 		ext4_commit_super(sb, 0);
802 		return;
803 	}
804 
805 	ext4_unlock_group(sb, grp);
806 	ext4_commit_super(sb, 1);
807 	ext4_handle_error(sb);
808 	/*
809 	 * We only get here in the ERRORS_RO case; relocking the group
810 	 * may be dangerous, but nothing bad will happen since the
811 	 * filesystem will have already been marked read/only and the
812 	 * journal has been aborted.  We return 1 as a hint to callers
813 	 * who might what to use the return value from
814 	 * ext4_grp_locked_error() to distinguish between the
815 	 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
816 	 * aggressively from the ext4 function in question, with a
817 	 * more appropriate error code.
818 	 */
819 	ext4_lock_group(sb, grp);
820 	return;
821 }
822 
ext4_mark_group_bitmap_corrupted(struct super_block * sb,ext4_group_t group,unsigned int flags)823 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
824 				     ext4_group_t group,
825 				     unsigned int flags)
826 {
827 	struct ext4_sb_info *sbi = EXT4_SB(sb);
828 	struct ext4_group_info *grp = ext4_get_group_info(sb, group);
829 	struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
830 	int ret;
831 
832 	if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
833 		ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
834 					    &grp->bb_state);
835 		if (!ret)
836 			percpu_counter_sub(&sbi->s_freeclusters_counter,
837 					   grp->bb_free);
838 	}
839 
840 	if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
841 		ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
842 					    &grp->bb_state);
843 		if (!ret && gdp) {
844 			int count;
845 
846 			count = ext4_free_inodes_count(sb, gdp);
847 			percpu_counter_sub(&sbi->s_freeinodes_counter,
848 					   count);
849 		}
850 	}
851 }
852 
ext4_update_dynamic_rev(struct super_block * sb)853 void ext4_update_dynamic_rev(struct super_block *sb)
854 {
855 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
856 
857 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
858 		return;
859 
860 	ext4_warning(sb,
861 		     "updating to rev %d because of new feature flag, "
862 		     "running e2fsck is recommended",
863 		     EXT4_DYNAMIC_REV);
864 
865 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
866 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
867 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
868 	/* leave es->s_feature_*compat flags alone */
869 	/* es->s_uuid will be set by e2fsck if empty */
870 
871 	/*
872 	 * The rest of the superblock fields should be zero, and if not it
873 	 * means they are likely already in use, so leave them alone.  We
874 	 * can leave it up to e2fsck to clean up any inconsistencies there.
875 	 */
876 }
877 
878 /*
879  * Open the external journal device
880  */
ext4_blkdev_get(dev_t dev,struct super_block * sb)881 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
882 {
883 	struct block_device *bdev;
884 	char b[BDEVNAME_SIZE];
885 
886 	bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
887 	if (IS_ERR(bdev))
888 		goto fail;
889 	return bdev;
890 
891 fail:
892 	ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
893 			__bdevname(dev, b), PTR_ERR(bdev));
894 	return NULL;
895 }
896 
897 /*
898  * Release the journal device
899  */
ext4_blkdev_put(struct block_device * bdev)900 static void ext4_blkdev_put(struct block_device *bdev)
901 {
902 	blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
903 }
904 
ext4_blkdev_remove(struct ext4_sb_info * sbi)905 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
906 {
907 	struct block_device *bdev;
908 	bdev = sbi->s_journal_bdev;
909 	if (bdev) {
910 		/*
911 		 * Invalidate the journal device's buffers.  We don't want them
912 		 * floating about in memory - the physical journal device may
913 		 * hotswapped, and it breaks the `ro-after' testing code.
914 		 */
915 		invalidate_bdev(bdev);
916 		ext4_blkdev_put(bdev);
917 		sbi->s_journal_bdev = NULL;
918 	}
919 }
920 
orphan_list_entry(struct list_head * l)921 static inline struct inode *orphan_list_entry(struct list_head *l)
922 {
923 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
924 }
925 
dump_orphan_list(struct super_block * sb,struct ext4_sb_info * sbi)926 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
927 {
928 	struct list_head *l;
929 
930 	ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
931 		 le32_to_cpu(sbi->s_es->s_last_orphan));
932 
933 	printk(KERN_ERR "sb_info orphan list:\n");
934 	list_for_each(l, &sbi->s_orphan) {
935 		struct inode *inode = orphan_list_entry(l);
936 		printk(KERN_ERR "  "
937 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
938 		       inode->i_sb->s_id, inode->i_ino, inode,
939 		       inode->i_mode, inode->i_nlink,
940 		       NEXT_ORPHAN(inode));
941 	}
942 }
943 
944 #ifdef CONFIG_QUOTA
945 static int ext4_quota_off(struct super_block *sb, int type);
946 
ext4_quota_off_umount(struct super_block * sb)947 static inline void ext4_quota_off_umount(struct super_block *sb)
948 {
949 	int type;
950 
951 	/* Use our quota_off function to clear inode flags etc. */
952 	for (type = 0; type < EXT4_MAXQUOTAS; type++)
953 		ext4_quota_off(sb, type);
954 }
955 
956 /*
957  * This is a helper function which is used in the mount/remount
958  * codepaths (which holds s_umount) to fetch the quota file name.
959  */
get_qf_name(struct super_block * sb,struct ext4_sb_info * sbi,int type)960 static inline char *get_qf_name(struct super_block *sb,
961 				struct ext4_sb_info *sbi,
962 				int type)
963 {
964 	return rcu_dereference_protected(sbi->s_qf_names[type],
965 					 lockdep_is_held(&sb->s_umount));
966 }
967 #else
ext4_quota_off_umount(struct super_block * sb)968 static inline void ext4_quota_off_umount(struct super_block *sb)
969 {
970 }
971 #endif
972 
ext4_put_super(struct super_block * sb)973 static void ext4_put_super(struct super_block *sb)
974 {
975 	struct ext4_sb_info *sbi = EXT4_SB(sb);
976 	struct ext4_super_block *es = sbi->s_es;
977 	struct buffer_head **group_desc;
978 	struct flex_groups **flex_groups;
979 	int aborted = 0;
980 	int i, err;
981 
982 	ext4_unregister_li_request(sb);
983 	ext4_quota_off_umount(sb);
984 
985 	destroy_workqueue(sbi->rsv_conversion_wq);
986 
987 	if (sbi->s_journal) {
988 		aborted = is_journal_aborted(sbi->s_journal);
989 		err = jbd2_journal_destroy(sbi->s_journal);
990 		sbi->s_journal = NULL;
991 		if ((err < 0) && !aborted)
992 			ext4_abort(sb, "Couldn't clean up the journal");
993 	}
994 
995 	ext4_unregister_sysfs(sb);
996 	ext4_es_unregister_shrinker(sbi);
997 	del_timer_sync(&sbi->s_err_report);
998 	ext4_release_system_zone(sb);
999 	ext4_mb_release(sb);
1000 	ext4_ext_release(sb);
1001 
1002 	if (!sb_rdonly(sb) && !aborted) {
1003 		ext4_clear_feature_journal_needs_recovery(sb);
1004 		es->s_state = cpu_to_le16(sbi->s_mount_state);
1005 	}
1006 	if (!sb_rdonly(sb))
1007 		ext4_commit_super(sb, 1);
1008 
1009 	rcu_read_lock();
1010 	group_desc = rcu_dereference(sbi->s_group_desc);
1011 	for (i = 0; i < sbi->s_gdb_count; i++)
1012 		brelse(group_desc[i]);
1013 	kvfree(group_desc);
1014 	flex_groups = rcu_dereference(sbi->s_flex_groups);
1015 	if (flex_groups) {
1016 		for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1017 			kvfree(flex_groups[i]);
1018 		kvfree(flex_groups);
1019 	}
1020 	rcu_read_unlock();
1021 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
1022 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1023 	percpu_counter_destroy(&sbi->s_dirs_counter);
1024 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1025 	percpu_free_rwsem(&sbi->s_writepages_rwsem);
1026 #ifdef CONFIG_QUOTA
1027 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
1028 		kfree(get_qf_name(sb, sbi, i));
1029 #endif
1030 
1031 	/* Debugging code just in case the in-memory inode orphan list
1032 	 * isn't empty.  The on-disk one can be non-empty if we've
1033 	 * detected an error and taken the fs readonly, but the
1034 	 * in-memory list had better be clean by this point. */
1035 	if (!list_empty(&sbi->s_orphan))
1036 		dump_orphan_list(sb, sbi);
1037 	J_ASSERT(list_empty(&sbi->s_orphan));
1038 
1039 	sync_blockdev(sb->s_bdev);
1040 	invalidate_bdev(sb->s_bdev);
1041 	if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1042 		sync_blockdev(sbi->s_journal_bdev);
1043 		ext4_blkdev_remove(sbi);
1044 	}
1045 	if (sbi->s_ea_inode_cache) {
1046 		ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1047 		sbi->s_ea_inode_cache = NULL;
1048 	}
1049 	if (sbi->s_ea_block_cache) {
1050 		ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1051 		sbi->s_ea_block_cache = NULL;
1052 	}
1053 	if (sbi->s_mmp_tsk)
1054 		kthread_stop(sbi->s_mmp_tsk);
1055 	brelse(sbi->s_sbh);
1056 	sb->s_fs_info = NULL;
1057 	/*
1058 	 * Now that we are completely done shutting down the
1059 	 * superblock, we need to actually destroy the kobject.
1060 	 */
1061 	kobject_put(&sbi->s_kobj);
1062 	wait_for_completion(&sbi->s_kobj_unregister);
1063 	if (sbi->s_chksum_driver)
1064 		crypto_free_shash(sbi->s_chksum_driver);
1065 	kfree(sbi->s_blockgroup_lock);
1066 	fs_put_dax(sbi->s_daxdev);
1067 	kfree(sbi);
1068 }
1069 
1070 static struct kmem_cache *ext4_inode_cachep;
1071 
1072 /*
1073  * Called inside transaction, so use GFP_NOFS
1074  */
ext4_alloc_inode(struct super_block * sb)1075 static struct inode *ext4_alloc_inode(struct super_block *sb)
1076 {
1077 	struct ext4_inode_info *ei;
1078 
1079 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1080 	if (!ei)
1081 		return NULL;
1082 
1083 	inode_set_iversion(&ei->vfs_inode, 1);
1084 	ei->i_flags = 0;
1085 	spin_lock_init(&ei->i_raw_lock);
1086 	INIT_LIST_HEAD(&ei->i_prealloc_list);
1087 	spin_lock_init(&ei->i_prealloc_lock);
1088 	ext4_es_init_tree(&ei->i_es_tree);
1089 	rwlock_init(&ei->i_es_lock);
1090 	INIT_LIST_HEAD(&ei->i_es_list);
1091 	ei->i_es_all_nr = 0;
1092 	ei->i_es_shk_nr = 0;
1093 	ei->i_es_shrink_lblk = 0;
1094 	ei->i_reserved_data_blocks = 0;
1095 	ei->i_da_metadata_calc_len = 0;
1096 	ei->i_da_metadata_calc_last_lblock = 0;
1097 	spin_lock_init(&(ei->i_block_reservation_lock));
1098 #ifdef CONFIG_QUOTA
1099 	ei->i_reserved_quota = 0;
1100 	memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1101 #endif
1102 	ei->jinode = NULL;
1103 	INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1104 	spin_lock_init(&ei->i_completed_io_lock);
1105 	ei->i_sync_tid = 0;
1106 	ei->i_datasync_tid = 0;
1107 	atomic_set(&ei->i_unwritten, 0);
1108 	INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1109 	return &ei->vfs_inode;
1110 }
1111 
ext4_drop_inode(struct inode * inode)1112 static int ext4_drop_inode(struct inode *inode)
1113 {
1114 	int drop = generic_drop_inode(inode);
1115 
1116 	trace_ext4_drop_inode(inode, drop);
1117 	return drop;
1118 }
1119 
ext4_i_callback(struct rcu_head * head)1120 static void ext4_i_callback(struct rcu_head *head)
1121 {
1122 	struct inode *inode = container_of(head, struct inode, i_rcu);
1123 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1124 }
1125 
ext4_destroy_inode(struct inode * inode)1126 static void ext4_destroy_inode(struct inode *inode)
1127 {
1128 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1129 		ext4_msg(inode->i_sb, KERN_ERR,
1130 			 "Inode %lu (%p): orphan list check failed!",
1131 			 inode->i_ino, EXT4_I(inode));
1132 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1133 				EXT4_I(inode), sizeof(struct ext4_inode_info),
1134 				true);
1135 		dump_stack();
1136 	}
1137 	call_rcu(&inode->i_rcu, ext4_i_callback);
1138 }
1139 
init_once(void * foo)1140 static void init_once(void *foo)
1141 {
1142 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1143 
1144 	INIT_LIST_HEAD(&ei->i_orphan);
1145 	init_rwsem(&ei->xattr_sem);
1146 	init_rwsem(&ei->i_data_sem);
1147 	init_rwsem(&ei->i_mmap_sem);
1148 	inode_init_once(&ei->vfs_inode);
1149 }
1150 
init_inodecache(void)1151 static int __init init_inodecache(void)
1152 {
1153 	ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1154 				sizeof(struct ext4_inode_info), 0,
1155 				(SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1156 					SLAB_ACCOUNT),
1157 				offsetof(struct ext4_inode_info, i_data),
1158 				sizeof_field(struct ext4_inode_info, i_data),
1159 				init_once);
1160 	if (ext4_inode_cachep == NULL)
1161 		return -ENOMEM;
1162 	return 0;
1163 }
1164 
destroy_inodecache(void)1165 static void destroy_inodecache(void)
1166 {
1167 	/*
1168 	 * Make sure all delayed rcu free inodes are flushed before we
1169 	 * destroy cache.
1170 	 */
1171 	rcu_barrier();
1172 	kmem_cache_destroy(ext4_inode_cachep);
1173 }
1174 
ext4_clear_inode(struct inode * inode)1175 void ext4_clear_inode(struct inode *inode)
1176 {
1177 	invalidate_inode_buffers(inode);
1178 	clear_inode(inode);
1179 	dquot_drop(inode);
1180 	ext4_discard_preallocations(inode);
1181 	ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1182 	if (EXT4_I(inode)->jinode) {
1183 		jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1184 					       EXT4_I(inode)->jinode);
1185 		jbd2_free_inode(EXT4_I(inode)->jinode);
1186 		EXT4_I(inode)->jinode = NULL;
1187 	}
1188 	fscrypt_put_encryption_info(inode);
1189 }
1190 
ext4_nfs_get_inode(struct super_block * sb,u64 ino,u32 generation)1191 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1192 					u64 ino, u32 generation)
1193 {
1194 	struct inode *inode;
1195 
1196 	/*
1197 	 * Currently we don't know the generation for parent directory, so
1198 	 * a generation of 0 means "accept any"
1199 	 */
1200 	inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1201 	if (IS_ERR(inode))
1202 		return ERR_CAST(inode);
1203 	if (generation && inode->i_generation != generation) {
1204 		iput(inode);
1205 		return ERR_PTR(-ESTALE);
1206 	}
1207 
1208 	return inode;
1209 }
1210 
ext4_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)1211 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1212 					int fh_len, int fh_type)
1213 {
1214 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1215 				    ext4_nfs_get_inode);
1216 }
1217 
ext4_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)1218 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1219 					int fh_len, int fh_type)
1220 {
1221 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1222 				    ext4_nfs_get_inode);
1223 }
1224 
ext4_nfs_commit_metadata(struct inode * inode)1225 static int ext4_nfs_commit_metadata(struct inode *inode)
1226 {
1227 	struct writeback_control wbc = {
1228 		.sync_mode = WB_SYNC_ALL
1229 	};
1230 
1231 	trace_ext4_nfs_commit_metadata(inode);
1232 	return ext4_write_inode(inode, &wbc);
1233 }
1234 
1235 /*
1236  * Try to release metadata pages (indirect blocks, directories) which are
1237  * mapped via the block device.  Since these pages could have journal heads
1238  * which would prevent try_to_free_buffers() from freeing them, we must use
1239  * jbd2 layer's try_to_free_buffers() function to release them.
1240  */
bdev_try_to_free_page(struct super_block * sb,struct page * page,gfp_t wait)1241 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1242 				 gfp_t wait)
1243 {
1244 	journal_t *journal = EXT4_SB(sb)->s_journal;
1245 
1246 	WARN_ON(PageChecked(page));
1247 	if (!page_has_buffers(page))
1248 		return 0;
1249 	if (journal)
1250 		return jbd2_journal_try_to_free_buffers(journal, page,
1251 						wait & ~__GFP_DIRECT_RECLAIM);
1252 	return try_to_free_buffers(page);
1253 }
1254 
1255 #ifdef CONFIG_EXT4_FS_ENCRYPTION
ext4_get_context(struct inode * inode,void * ctx,size_t len)1256 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1257 {
1258 	return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1259 				 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1260 }
1261 
ext4_set_context(struct inode * inode,const void * ctx,size_t len,void * fs_data)1262 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1263 							void *fs_data)
1264 {
1265 	handle_t *handle = fs_data;
1266 	int res, res2, credits, retries = 0;
1267 
1268 	/*
1269 	 * Encrypting the root directory is not allowed because e2fsck expects
1270 	 * lost+found to exist and be unencrypted, and encrypting the root
1271 	 * directory would imply encrypting the lost+found directory as well as
1272 	 * the filename "lost+found" itself.
1273 	 */
1274 	if (inode->i_ino == EXT4_ROOT_INO)
1275 		return -EPERM;
1276 
1277 	if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1278 		return -EINVAL;
1279 
1280 	res = ext4_convert_inline_data(inode);
1281 	if (res)
1282 		return res;
1283 
1284 	/*
1285 	 * If a journal handle was specified, then the encryption context is
1286 	 * being set on a new inode via inheritance and is part of a larger
1287 	 * transaction to create the inode.  Otherwise the encryption context is
1288 	 * being set on an existing inode in its own transaction.  Only in the
1289 	 * latter case should the "retry on ENOSPC" logic be used.
1290 	 */
1291 
1292 	if (handle) {
1293 		res = ext4_xattr_set_handle(handle, inode,
1294 					    EXT4_XATTR_INDEX_ENCRYPTION,
1295 					    EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1296 					    ctx, len, 0);
1297 		if (!res) {
1298 			ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1299 			ext4_clear_inode_state(inode,
1300 					EXT4_STATE_MAY_INLINE_DATA);
1301 			/*
1302 			 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1303 			 * S_DAX may be disabled
1304 			 */
1305 			ext4_set_inode_flags(inode);
1306 		}
1307 		return res;
1308 	}
1309 
1310 	res = dquot_initialize(inode);
1311 	if (res)
1312 		return res;
1313 retry:
1314 	res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1315 				     &credits);
1316 	if (res)
1317 		return res;
1318 
1319 	handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1320 	if (IS_ERR(handle))
1321 		return PTR_ERR(handle);
1322 
1323 	res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1324 				    EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1325 				    ctx, len, 0);
1326 	if (!res) {
1327 		ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1328 		/*
1329 		 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1330 		 * S_DAX may be disabled
1331 		 */
1332 		ext4_set_inode_flags(inode);
1333 		res = ext4_mark_inode_dirty(handle, inode);
1334 		if (res)
1335 			EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1336 	}
1337 	res2 = ext4_journal_stop(handle);
1338 
1339 	if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1340 		goto retry;
1341 	if (!res)
1342 		res = res2;
1343 	return res;
1344 }
1345 
ext4_dummy_context(struct inode * inode)1346 static bool ext4_dummy_context(struct inode *inode)
1347 {
1348 	return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1349 }
1350 
1351 static const struct fscrypt_operations ext4_cryptops = {
1352 	.key_prefix		= "ext4:",
1353 	.get_context		= ext4_get_context,
1354 	.set_context		= ext4_set_context,
1355 	.dummy_context		= ext4_dummy_context,
1356 	.empty_dir		= ext4_empty_dir,
1357 	.max_namelen		= EXT4_NAME_LEN,
1358 };
1359 #endif
1360 
1361 #ifdef CONFIG_QUOTA
1362 static const char * const quotatypes[] = INITQFNAMES;
1363 #define QTYPE2NAME(t) (quotatypes[t])
1364 
1365 static int ext4_write_dquot(struct dquot *dquot);
1366 static int ext4_acquire_dquot(struct dquot *dquot);
1367 static int ext4_release_dquot(struct dquot *dquot);
1368 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1369 static int ext4_write_info(struct super_block *sb, int type);
1370 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1371 			 const struct path *path);
1372 static int ext4_quota_on_mount(struct super_block *sb, int type);
1373 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1374 			       size_t len, loff_t off);
1375 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1376 				const char *data, size_t len, loff_t off);
1377 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1378 			     unsigned int flags);
1379 static int ext4_enable_quotas(struct super_block *sb);
1380 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1381 
ext4_get_dquots(struct inode * inode)1382 static struct dquot **ext4_get_dquots(struct inode *inode)
1383 {
1384 	return EXT4_I(inode)->i_dquot;
1385 }
1386 
1387 static const struct dquot_operations ext4_quota_operations = {
1388 	.get_reserved_space	= ext4_get_reserved_space,
1389 	.write_dquot		= ext4_write_dquot,
1390 	.acquire_dquot		= ext4_acquire_dquot,
1391 	.release_dquot		= ext4_release_dquot,
1392 	.mark_dirty		= ext4_mark_dquot_dirty,
1393 	.write_info		= ext4_write_info,
1394 	.alloc_dquot		= dquot_alloc,
1395 	.destroy_dquot		= dquot_destroy,
1396 	.get_projid		= ext4_get_projid,
1397 	.get_inode_usage	= ext4_get_inode_usage,
1398 	.get_next_id		= ext4_get_next_id,
1399 };
1400 
1401 static const struct quotactl_ops ext4_qctl_operations = {
1402 	.quota_on	= ext4_quota_on,
1403 	.quota_off	= ext4_quota_off,
1404 	.quota_sync	= dquot_quota_sync,
1405 	.get_state	= dquot_get_state,
1406 	.set_info	= dquot_set_dqinfo,
1407 	.get_dqblk	= dquot_get_dqblk,
1408 	.set_dqblk	= dquot_set_dqblk,
1409 	.get_nextdqblk	= dquot_get_next_dqblk,
1410 };
1411 #endif
1412 
1413 static const struct super_operations ext4_sops = {
1414 	.alloc_inode	= ext4_alloc_inode,
1415 	.destroy_inode	= ext4_destroy_inode,
1416 	.write_inode	= ext4_write_inode,
1417 	.dirty_inode	= ext4_dirty_inode,
1418 	.drop_inode	= ext4_drop_inode,
1419 	.evict_inode	= ext4_evict_inode,
1420 	.put_super	= ext4_put_super,
1421 	.sync_fs	= ext4_sync_fs,
1422 	.freeze_fs	= ext4_freeze,
1423 	.unfreeze_fs	= ext4_unfreeze,
1424 	.statfs		= ext4_statfs,
1425 	.remount_fs	= ext4_remount,
1426 	.show_options	= ext4_show_options,
1427 #ifdef CONFIG_QUOTA
1428 	.quota_read	= ext4_quota_read,
1429 	.quota_write	= ext4_quota_write,
1430 	.get_dquots	= ext4_get_dquots,
1431 #endif
1432 	.bdev_try_to_free_page = bdev_try_to_free_page,
1433 };
1434 
1435 static const struct export_operations ext4_export_ops = {
1436 	.fh_to_dentry = ext4_fh_to_dentry,
1437 	.fh_to_parent = ext4_fh_to_parent,
1438 	.get_parent = ext4_get_parent,
1439 	.commit_metadata = ext4_nfs_commit_metadata,
1440 };
1441 
1442 enum {
1443 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1444 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1445 	Opt_nouid32, Opt_debug, Opt_removed,
1446 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1447 	Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1448 	Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1449 	Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1450 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1451 	Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1452 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1453 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1454 	Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1455 	Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1456 	Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1457 	Opt_nowarn_on_error, Opt_mblk_io_submit,
1458 	Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1459 	Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1460 	Opt_inode_readahead_blks, Opt_journal_ioprio,
1461 	Opt_dioread_nolock, Opt_dioread_lock,
1462 	Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1463 	Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1464 };
1465 
1466 static const match_table_t tokens = {
1467 	{Opt_bsd_df, "bsddf"},
1468 	{Opt_minix_df, "minixdf"},
1469 	{Opt_grpid, "grpid"},
1470 	{Opt_grpid, "bsdgroups"},
1471 	{Opt_nogrpid, "nogrpid"},
1472 	{Opt_nogrpid, "sysvgroups"},
1473 	{Opt_resgid, "resgid=%u"},
1474 	{Opt_resuid, "resuid=%u"},
1475 	{Opt_sb, "sb=%u"},
1476 	{Opt_err_cont, "errors=continue"},
1477 	{Opt_err_panic, "errors=panic"},
1478 	{Opt_err_ro, "errors=remount-ro"},
1479 	{Opt_nouid32, "nouid32"},
1480 	{Opt_debug, "debug"},
1481 	{Opt_removed, "oldalloc"},
1482 	{Opt_removed, "orlov"},
1483 	{Opt_user_xattr, "user_xattr"},
1484 	{Opt_nouser_xattr, "nouser_xattr"},
1485 	{Opt_acl, "acl"},
1486 	{Opt_noacl, "noacl"},
1487 	{Opt_noload, "norecovery"},
1488 	{Opt_noload, "noload"},
1489 	{Opt_removed, "nobh"},
1490 	{Opt_removed, "bh"},
1491 	{Opt_commit, "commit=%u"},
1492 	{Opt_min_batch_time, "min_batch_time=%u"},
1493 	{Opt_max_batch_time, "max_batch_time=%u"},
1494 	{Opt_journal_dev, "journal_dev=%u"},
1495 	{Opt_journal_path, "journal_path=%s"},
1496 	{Opt_journal_checksum, "journal_checksum"},
1497 	{Opt_nojournal_checksum, "nojournal_checksum"},
1498 	{Opt_journal_async_commit, "journal_async_commit"},
1499 	{Opt_abort, "abort"},
1500 	{Opt_data_journal, "data=journal"},
1501 	{Opt_data_ordered, "data=ordered"},
1502 	{Opt_data_writeback, "data=writeback"},
1503 	{Opt_data_err_abort, "data_err=abort"},
1504 	{Opt_data_err_ignore, "data_err=ignore"},
1505 	{Opt_offusrjquota, "usrjquota="},
1506 	{Opt_usrjquota, "usrjquota=%s"},
1507 	{Opt_offgrpjquota, "grpjquota="},
1508 	{Opt_grpjquota, "grpjquota=%s"},
1509 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1510 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1511 	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1512 	{Opt_grpquota, "grpquota"},
1513 	{Opt_noquota, "noquota"},
1514 	{Opt_quota, "quota"},
1515 	{Opt_usrquota, "usrquota"},
1516 	{Opt_prjquota, "prjquota"},
1517 	{Opt_barrier, "barrier=%u"},
1518 	{Opt_barrier, "barrier"},
1519 	{Opt_nobarrier, "nobarrier"},
1520 	{Opt_i_version, "i_version"},
1521 	{Opt_dax, "dax"},
1522 	{Opt_stripe, "stripe=%u"},
1523 	{Opt_delalloc, "delalloc"},
1524 	{Opt_warn_on_error, "warn_on_error"},
1525 	{Opt_nowarn_on_error, "nowarn_on_error"},
1526 	{Opt_lazytime, "lazytime"},
1527 	{Opt_nolazytime, "nolazytime"},
1528 	{Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1529 	{Opt_nodelalloc, "nodelalloc"},
1530 	{Opt_removed, "mblk_io_submit"},
1531 	{Opt_removed, "nomblk_io_submit"},
1532 	{Opt_block_validity, "block_validity"},
1533 	{Opt_noblock_validity, "noblock_validity"},
1534 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1535 	{Opt_journal_ioprio, "journal_ioprio=%u"},
1536 	{Opt_auto_da_alloc, "auto_da_alloc=%u"},
1537 	{Opt_auto_da_alloc, "auto_da_alloc"},
1538 	{Opt_noauto_da_alloc, "noauto_da_alloc"},
1539 	{Opt_dioread_nolock, "dioread_nolock"},
1540 	{Opt_dioread_lock, "dioread_lock"},
1541 	{Opt_discard, "discard"},
1542 	{Opt_nodiscard, "nodiscard"},
1543 	{Opt_init_itable, "init_itable=%u"},
1544 	{Opt_init_itable, "init_itable"},
1545 	{Opt_noinit_itable, "noinit_itable"},
1546 	{Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1547 	{Opt_test_dummy_encryption, "test_dummy_encryption"},
1548 	{Opt_nombcache, "nombcache"},
1549 	{Opt_nombcache, "no_mbcache"},	/* for backward compatibility */
1550 	{Opt_removed, "check=none"},	/* mount option from ext2/3 */
1551 	{Opt_removed, "nocheck"},	/* mount option from ext2/3 */
1552 	{Opt_removed, "reservation"},	/* mount option from ext2/3 */
1553 	{Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1554 	{Opt_removed, "journal=%u"},	/* mount option from ext2/3 */
1555 	{Opt_err, NULL},
1556 };
1557 
get_sb_block(void ** data)1558 static ext4_fsblk_t get_sb_block(void **data)
1559 {
1560 	ext4_fsblk_t	sb_block;
1561 	char		*options = (char *) *data;
1562 
1563 	if (!options || strncmp(options, "sb=", 3) != 0)
1564 		return 1;	/* Default location */
1565 
1566 	options += 3;
1567 	/* TODO: use simple_strtoll with >32bit ext4 */
1568 	sb_block = simple_strtoul(options, &options, 0);
1569 	if (*options && *options != ',') {
1570 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1571 		       (char *) *data);
1572 		return 1;
1573 	}
1574 	if (*options == ',')
1575 		options++;
1576 	*data = (void *) options;
1577 
1578 	return sb_block;
1579 }
1580 
1581 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1582 static const char deprecated_msg[] =
1583 	"Mount option \"%s\" will be removed by %s\n"
1584 	"Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1585 
1586 #ifdef CONFIG_QUOTA
set_qf_name(struct super_block * sb,int qtype,substring_t * args)1587 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1588 {
1589 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1590 	char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1591 	int ret = -1;
1592 
1593 	if (sb_any_quota_loaded(sb) && !old_qname) {
1594 		ext4_msg(sb, KERN_ERR,
1595 			"Cannot change journaled "
1596 			"quota options when quota turned on");
1597 		return -1;
1598 	}
1599 	if (ext4_has_feature_quota(sb)) {
1600 		ext4_msg(sb, KERN_INFO, "Journaled quota options "
1601 			 "ignored when QUOTA feature is enabled");
1602 		return 1;
1603 	}
1604 	qname = match_strdup(args);
1605 	if (!qname) {
1606 		ext4_msg(sb, KERN_ERR,
1607 			"Not enough memory for storing quotafile name");
1608 		return -1;
1609 	}
1610 	if (old_qname) {
1611 		if (strcmp(old_qname, qname) == 0)
1612 			ret = 1;
1613 		else
1614 			ext4_msg(sb, KERN_ERR,
1615 				 "%s quota file already specified",
1616 				 QTYPE2NAME(qtype));
1617 		goto errout;
1618 	}
1619 	if (strchr(qname, '/')) {
1620 		ext4_msg(sb, KERN_ERR,
1621 			"quotafile must be on filesystem root");
1622 		goto errout;
1623 	}
1624 	rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1625 	set_opt(sb, QUOTA);
1626 	return 1;
1627 errout:
1628 	kfree(qname);
1629 	return ret;
1630 }
1631 
clear_qf_name(struct super_block * sb,int qtype)1632 static int clear_qf_name(struct super_block *sb, int qtype)
1633 {
1634 
1635 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1636 	char *old_qname = get_qf_name(sb, sbi, qtype);
1637 
1638 	if (sb_any_quota_loaded(sb) && old_qname) {
1639 		ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1640 			" when quota turned on");
1641 		return -1;
1642 	}
1643 	rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1644 	synchronize_rcu();
1645 	kfree(old_qname);
1646 	return 1;
1647 }
1648 #endif
1649 
1650 #define MOPT_SET	0x0001
1651 #define MOPT_CLEAR	0x0002
1652 #define MOPT_NOSUPPORT	0x0004
1653 #define MOPT_EXPLICIT	0x0008
1654 #define MOPT_CLEAR_ERR	0x0010
1655 #define MOPT_GTE0	0x0020
1656 #ifdef CONFIG_QUOTA
1657 #define MOPT_Q		0
1658 #define MOPT_QFMT	0x0040
1659 #else
1660 #define MOPT_Q		MOPT_NOSUPPORT
1661 #define MOPT_QFMT	MOPT_NOSUPPORT
1662 #endif
1663 #define MOPT_DATAJ	0x0080
1664 #define MOPT_NO_EXT2	0x0100
1665 #define MOPT_NO_EXT3	0x0200
1666 #define MOPT_EXT4_ONLY	(MOPT_NO_EXT2 | MOPT_NO_EXT3)
1667 #define MOPT_STRING	0x0400
1668 
1669 static const struct mount_opts {
1670 	int	token;
1671 	int	mount_opt;
1672 	int	flags;
1673 } ext4_mount_opts[] = {
1674 	{Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1675 	{Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1676 	{Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1677 	{Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1678 	{Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1679 	{Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1680 	{Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1681 	 MOPT_EXT4_ONLY | MOPT_SET},
1682 	{Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1683 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1684 	{Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1685 	{Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1686 	{Opt_delalloc, EXT4_MOUNT_DELALLOC,
1687 	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1688 	{Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1689 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1690 	{Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1691 	{Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1692 	{Opt_commit, 0, MOPT_NO_EXT2},
1693 	{Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1694 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1695 	{Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1696 	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1697 	{Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1698 				    EXT4_MOUNT_JOURNAL_CHECKSUM),
1699 	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1700 	{Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1701 	{Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1702 	{Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1703 	{Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1704 	{Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1705 	 MOPT_NO_EXT2},
1706 	{Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1707 	 MOPT_NO_EXT2},
1708 	{Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1709 	{Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1710 	{Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1711 	{Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1712 	{Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1713 	{Opt_commit, 0, MOPT_GTE0},
1714 	{Opt_max_batch_time, 0, MOPT_GTE0},
1715 	{Opt_min_batch_time, 0, MOPT_GTE0},
1716 	{Opt_inode_readahead_blks, 0, MOPT_GTE0},
1717 	{Opt_init_itable, 0, MOPT_GTE0},
1718 	{Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1719 	{Opt_stripe, 0, MOPT_GTE0},
1720 	{Opt_resuid, 0, MOPT_GTE0},
1721 	{Opt_resgid, 0, MOPT_GTE0},
1722 	{Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1723 	{Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1724 	{Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1725 	{Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1726 	{Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1727 	{Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1728 	 MOPT_NO_EXT2 | MOPT_DATAJ},
1729 	{Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1730 	{Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1731 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1732 	{Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1733 	{Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1734 #else
1735 	{Opt_acl, 0, MOPT_NOSUPPORT},
1736 	{Opt_noacl, 0, MOPT_NOSUPPORT},
1737 #endif
1738 	{Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1739 	{Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1740 	{Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1741 	{Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1742 	{Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1743 							MOPT_SET | MOPT_Q},
1744 	{Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1745 							MOPT_SET | MOPT_Q},
1746 	{Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1747 							MOPT_SET | MOPT_Q},
1748 	{Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1749 		       EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1750 							MOPT_CLEAR | MOPT_Q},
1751 	{Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
1752 	{Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
1753 	{Opt_offusrjquota, 0, MOPT_Q},
1754 	{Opt_offgrpjquota, 0, MOPT_Q},
1755 	{Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1756 	{Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1757 	{Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1758 	{Opt_max_dir_size_kb, 0, MOPT_GTE0},
1759 	{Opt_test_dummy_encryption, 0, MOPT_GTE0},
1760 	{Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1761 	{Opt_err, 0, 0}
1762 };
1763 
handle_mount_opt(struct super_block * sb,char * opt,int token,substring_t * args,unsigned long * journal_devnum,unsigned int * journal_ioprio,int is_remount)1764 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1765 			    substring_t *args, unsigned long *journal_devnum,
1766 			    unsigned int *journal_ioprio, int is_remount)
1767 {
1768 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1769 	const struct mount_opts *m;
1770 	kuid_t uid;
1771 	kgid_t gid;
1772 	int arg = 0;
1773 
1774 #ifdef CONFIG_QUOTA
1775 	if (token == Opt_usrjquota)
1776 		return set_qf_name(sb, USRQUOTA, &args[0]);
1777 	else if (token == Opt_grpjquota)
1778 		return set_qf_name(sb, GRPQUOTA, &args[0]);
1779 	else if (token == Opt_offusrjquota)
1780 		return clear_qf_name(sb, USRQUOTA);
1781 	else if (token == Opt_offgrpjquota)
1782 		return clear_qf_name(sb, GRPQUOTA);
1783 #endif
1784 	switch (token) {
1785 	case Opt_noacl:
1786 	case Opt_nouser_xattr:
1787 		ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1788 		break;
1789 	case Opt_sb:
1790 		return 1;	/* handled by get_sb_block() */
1791 	case Opt_removed:
1792 		ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1793 		return 1;
1794 	case Opt_abort:
1795 		sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1796 		return 1;
1797 	case Opt_i_version:
1798 		sb->s_flags |= SB_I_VERSION;
1799 		return 1;
1800 	case Opt_lazytime:
1801 		sb->s_flags |= SB_LAZYTIME;
1802 		return 1;
1803 	case Opt_nolazytime:
1804 		sb->s_flags &= ~SB_LAZYTIME;
1805 		return 1;
1806 	}
1807 
1808 	for (m = ext4_mount_opts; m->token != Opt_err; m++)
1809 		if (token == m->token)
1810 			break;
1811 
1812 	if (m->token == Opt_err) {
1813 		ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1814 			 "or missing value", opt);
1815 		return -1;
1816 	}
1817 
1818 	if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1819 		ext4_msg(sb, KERN_ERR,
1820 			 "Mount option \"%s\" incompatible with ext2", opt);
1821 		return -1;
1822 	}
1823 	if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1824 		ext4_msg(sb, KERN_ERR,
1825 			 "Mount option \"%s\" incompatible with ext3", opt);
1826 		return -1;
1827 	}
1828 
1829 	if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1830 		return -1;
1831 	if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1832 		return -1;
1833 	if (m->flags & MOPT_EXPLICIT) {
1834 		if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1835 			set_opt2(sb, EXPLICIT_DELALLOC);
1836 		} else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1837 			set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1838 		} else
1839 			return -1;
1840 	}
1841 	if (m->flags & MOPT_CLEAR_ERR)
1842 		clear_opt(sb, ERRORS_MASK);
1843 	if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1844 		ext4_msg(sb, KERN_ERR, "Cannot change quota "
1845 			 "options when quota turned on");
1846 		return -1;
1847 	}
1848 
1849 	if (m->flags & MOPT_NOSUPPORT) {
1850 		ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1851 	} else if (token == Opt_commit) {
1852 		if (arg == 0)
1853 			arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1854 		sbi->s_commit_interval = HZ * arg;
1855 	} else if (token == Opt_debug_want_extra_isize) {
1856 		if ((arg & 1) ||
1857 		    (arg < 4) ||
1858 		    (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
1859 			ext4_msg(sb, KERN_ERR,
1860 				 "Invalid want_extra_isize %d", arg);
1861 			return -1;
1862 		}
1863 		sbi->s_want_extra_isize = arg;
1864 	} else if (token == Opt_max_batch_time) {
1865 		sbi->s_max_batch_time = arg;
1866 	} else if (token == Opt_min_batch_time) {
1867 		sbi->s_min_batch_time = arg;
1868 	} else if (token == Opt_inode_readahead_blks) {
1869 		if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1870 			ext4_msg(sb, KERN_ERR,
1871 				 "EXT4-fs: inode_readahead_blks must be "
1872 				 "0 or a power of 2 smaller than 2^31");
1873 			return -1;
1874 		}
1875 		sbi->s_inode_readahead_blks = arg;
1876 	} else if (token == Opt_init_itable) {
1877 		set_opt(sb, INIT_INODE_TABLE);
1878 		if (!args->from)
1879 			arg = EXT4_DEF_LI_WAIT_MULT;
1880 		sbi->s_li_wait_mult = arg;
1881 	} else if (token == Opt_max_dir_size_kb) {
1882 		sbi->s_max_dir_size_kb = arg;
1883 	} else if (token == Opt_stripe) {
1884 		sbi->s_stripe = arg;
1885 	} else if (token == Opt_resuid) {
1886 		uid = make_kuid(current_user_ns(), arg);
1887 		if (!uid_valid(uid)) {
1888 			ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1889 			return -1;
1890 		}
1891 		sbi->s_resuid = uid;
1892 	} else if (token == Opt_resgid) {
1893 		gid = make_kgid(current_user_ns(), arg);
1894 		if (!gid_valid(gid)) {
1895 			ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1896 			return -1;
1897 		}
1898 		sbi->s_resgid = gid;
1899 	} else if (token == Opt_journal_dev) {
1900 		if (is_remount) {
1901 			ext4_msg(sb, KERN_ERR,
1902 				 "Cannot specify journal on remount");
1903 			return -1;
1904 		}
1905 		*journal_devnum = arg;
1906 	} else if (token == Opt_journal_path) {
1907 		char *journal_path;
1908 		struct inode *journal_inode;
1909 		struct path path;
1910 		int error;
1911 
1912 		if (is_remount) {
1913 			ext4_msg(sb, KERN_ERR,
1914 				 "Cannot specify journal on remount");
1915 			return -1;
1916 		}
1917 		journal_path = match_strdup(&args[0]);
1918 		if (!journal_path) {
1919 			ext4_msg(sb, KERN_ERR, "error: could not dup "
1920 				"journal device string");
1921 			return -1;
1922 		}
1923 
1924 		error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1925 		if (error) {
1926 			ext4_msg(sb, KERN_ERR, "error: could not find "
1927 				"journal device path: error %d", error);
1928 			kfree(journal_path);
1929 			return -1;
1930 		}
1931 
1932 		journal_inode = d_inode(path.dentry);
1933 		if (!S_ISBLK(journal_inode->i_mode)) {
1934 			ext4_msg(sb, KERN_ERR, "error: journal path %s "
1935 				"is not a block device", journal_path);
1936 			path_put(&path);
1937 			kfree(journal_path);
1938 			return -1;
1939 		}
1940 
1941 		*journal_devnum = new_encode_dev(journal_inode->i_rdev);
1942 		path_put(&path);
1943 		kfree(journal_path);
1944 	} else if (token == Opt_journal_ioprio) {
1945 		if (arg > 7) {
1946 			ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1947 				 " (must be 0-7)");
1948 			return -1;
1949 		}
1950 		*journal_ioprio =
1951 			IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1952 	} else if (token == Opt_test_dummy_encryption) {
1953 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1954 		sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1955 		ext4_msg(sb, KERN_WARNING,
1956 			 "Test dummy encryption mode enabled");
1957 #else
1958 		ext4_msg(sb, KERN_WARNING,
1959 			 "Test dummy encryption mount option ignored");
1960 #endif
1961 	} else if (m->flags & MOPT_DATAJ) {
1962 		if (is_remount) {
1963 			if (!sbi->s_journal)
1964 				ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1965 			else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1966 				ext4_msg(sb, KERN_ERR,
1967 					 "Cannot change data mode on remount");
1968 				return -1;
1969 			}
1970 		} else {
1971 			clear_opt(sb, DATA_FLAGS);
1972 			sbi->s_mount_opt |= m->mount_opt;
1973 		}
1974 #ifdef CONFIG_QUOTA
1975 	} else if (m->flags & MOPT_QFMT) {
1976 		if (sb_any_quota_loaded(sb) &&
1977 		    sbi->s_jquota_fmt != m->mount_opt) {
1978 			ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1979 				 "quota options when quota turned on");
1980 			return -1;
1981 		}
1982 		if (ext4_has_feature_quota(sb)) {
1983 			ext4_msg(sb, KERN_INFO,
1984 				 "Quota format mount options ignored "
1985 				 "when QUOTA feature is enabled");
1986 			return 1;
1987 		}
1988 		sbi->s_jquota_fmt = m->mount_opt;
1989 #endif
1990 	} else if (token == Opt_dax) {
1991 #ifdef CONFIG_FS_DAX
1992 		if (is_remount && test_opt(sb, DAX)) {
1993 			ext4_msg(sb, KERN_ERR, "can't mount with "
1994 				"both data=journal and dax");
1995 			return -1;
1996 		}
1997 		if (is_remount && !(sbi->s_mount_opt & EXT4_MOUNT_DAX)) {
1998 			ext4_msg(sb, KERN_ERR, "can't change "
1999 					"dax mount option while remounting");
2000 			return -1;
2001 		}
2002 		ext4_msg(sb, KERN_WARNING,
2003 		"DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2004 			sbi->s_mount_opt |= m->mount_opt;
2005 #else
2006 		ext4_msg(sb, KERN_INFO, "dax option not supported");
2007 		return -1;
2008 #endif
2009 	} else if (token == Opt_data_err_abort) {
2010 		sbi->s_mount_opt |= m->mount_opt;
2011 	} else if (token == Opt_data_err_ignore) {
2012 		sbi->s_mount_opt &= ~m->mount_opt;
2013 	} else {
2014 		if (!args->from)
2015 			arg = 1;
2016 		if (m->flags & MOPT_CLEAR)
2017 			arg = !arg;
2018 		else if (unlikely(!(m->flags & MOPT_SET))) {
2019 			ext4_msg(sb, KERN_WARNING,
2020 				 "buggy handling of option %s", opt);
2021 			WARN_ON(1);
2022 			return -1;
2023 		}
2024 		if (arg != 0)
2025 			sbi->s_mount_opt |= m->mount_opt;
2026 		else
2027 			sbi->s_mount_opt &= ~m->mount_opt;
2028 	}
2029 	return 1;
2030 }
2031 
parse_options(char * options,struct super_block * sb,unsigned long * journal_devnum,unsigned int * journal_ioprio,int is_remount)2032 static int parse_options(char *options, struct super_block *sb,
2033 			 unsigned long *journal_devnum,
2034 			 unsigned int *journal_ioprio,
2035 			 int is_remount)
2036 {
2037 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2038 	char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2039 	substring_t args[MAX_OPT_ARGS];
2040 	int token;
2041 
2042 	if (!options)
2043 		return 1;
2044 
2045 	while ((p = strsep(&options, ",")) != NULL) {
2046 		if (!*p)
2047 			continue;
2048 		/*
2049 		 * Initialize args struct so we know whether arg was
2050 		 * found; some options take optional arguments.
2051 		 */
2052 		args[0].to = args[0].from = NULL;
2053 		token = match_token(p, tokens, args);
2054 		if (handle_mount_opt(sb, p, token, args, journal_devnum,
2055 				     journal_ioprio, is_remount) < 0)
2056 			return 0;
2057 	}
2058 #ifdef CONFIG_QUOTA
2059 	/*
2060 	 * We do the test below only for project quotas. 'usrquota' and
2061 	 * 'grpquota' mount options are allowed even without quota feature
2062 	 * to support legacy quotas in quota files.
2063 	 */
2064 	if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2065 		ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2066 			 "Cannot enable project quota enforcement.");
2067 		return 0;
2068 	}
2069 	usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2070 	grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2071 	if (usr_qf_name || grp_qf_name) {
2072 		if (test_opt(sb, USRQUOTA) && usr_qf_name)
2073 			clear_opt(sb, USRQUOTA);
2074 
2075 		if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2076 			clear_opt(sb, GRPQUOTA);
2077 
2078 		if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2079 			ext4_msg(sb, KERN_ERR, "old and new quota "
2080 					"format mixing");
2081 			return 0;
2082 		}
2083 
2084 		if (!sbi->s_jquota_fmt) {
2085 			ext4_msg(sb, KERN_ERR, "journaled quota format "
2086 					"not specified");
2087 			return 0;
2088 		}
2089 	}
2090 #endif
2091 	if (test_opt(sb, DIOREAD_NOLOCK)) {
2092 		int blocksize =
2093 			BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2094 
2095 		if (blocksize < PAGE_SIZE) {
2096 			ext4_msg(sb, KERN_ERR, "can't mount with "
2097 				 "dioread_nolock if block size != PAGE_SIZE");
2098 			return 0;
2099 		}
2100 	}
2101 	return 1;
2102 }
2103 
ext4_show_quota_options(struct seq_file * seq,struct super_block * sb)2104 static inline void ext4_show_quota_options(struct seq_file *seq,
2105 					   struct super_block *sb)
2106 {
2107 #if defined(CONFIG_QUOTA)
2108 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2109 	char *usr_qf_name, *grp_qf_name;
2110 
2111 	if (sbi->s_jquota_fmt) {
2112 		char *fmtname = "";
2113 
2114 		switch (sbi->s_jquota_fmt) {
2115 		case QFMT_VFS_OLD:
2116 			fmtname = "vfsold";
2117 			break;
2118 		case QFMT_VFS_V0:
2119 			fmtname = "vfsv0";
2120 			break;
2121 		case QFMT_VFS_V1:
2122 			fmtname = "vfsv1";
2123 			break;
2124 		}
2125 		seq_printf(seq, ",jqfmt=%s", fmtname);
2126 	}
2127 
2128 	rcu_read_lock();
2129 	usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2130 	grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2131 	if (usr_qf_name)
2132 		seq_show_option(seq, "usrjquota", usr_qf_name);
2133 	if (grp_qf_name)
2134 		seq_show_option(seq, "grpjquota", grp_qf_name);
2135 	rcu_read_unlock();
2136 #endif
2137 }
2138 
token2str(int token)2139 static const char *token2str(int token)
2140 {
2141 	const struct match_token *t;
2142 
2143 	for (t = tokens; t->token != Opt_err; t++)
2144 		if (t->token == token && !strchr(t->pattern, '='))
2145 			break;
2146 	return t->pattern;
2147 }
2148 
2149 /*
2150  * Show an option if
2151  *  - it's set to a non-default value OR
2152  *  - if the per-sb default is different from the global default
2153  */
_ext4_show_options(struct seq_file * seq,struct super_block * sb,int nodefs)2154 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2155 			      int nodefs)
2156 {
2157 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2158 	struct ext4_super_block *es = sbi->s_es;
2159 	int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2160 	const struct mount_opts *m;
2161 	char sep = nodefs ? '\n' : ',';
2162 
2163 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2164 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2165 
2166 	if (sbi->s_sb_block != 1)
2167 		SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2168 
2169 	for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2170 		int want_set = m->flags & MOPT_SET;
2171 		if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2172 		    (m->flags & MOPT_CLEAR_ERR))
2173 			continue;
2174 		if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2175 			continue; /* skip if same as the default */
2176 		if ((want_set &&
2177 		     (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2178 		    (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2179 			continue; /* select Opt_noFoo vs Opt_Foo */
2180 		SEQ_OPTS_PRINT("%s", token2str(m->token));
2181 	}
2182 
2183 	if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2184 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2185 		SEQ_OPTS_PRINT("resuid=%u",
2186 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
2187 	if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2188 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2189 		SEQ_OPTS_PRINT("resgid=%u",
2190 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
2191 	def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2192 	if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2193 		SEQ_OPTS_PUTS("errors=remount-ro");
2194 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2195 		SEQ_OPTS_PUTS("errors=continue");
2196 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2197 		SEQ_OPTS_PUTS("errors=panic");
2198 	if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2199 		SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2200 	if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2201 		SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2202 	if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2203 		SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2204 	if (sb->s_flags & SB_I_VERSION)
2205 		SEQ_OPTS_PUTS("i_version");
2206 	if (nodefs || sbi->s_stripe)
2207 		SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2208 	if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2209 			(sbi->s_mount_opt ^ def_mount_opt)) {
2210 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2211 			SEQ_OPTS_PUTS("data=journal");
2212 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2213 			SEQ_OPTS_PUTS("data=ordered");
2214 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2215 			SEQ_OPTS_PUTS("data=writeback");
2216 	}
2217 	if (nodefs ||
2218 	    sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2219 		SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2220 			       sbi->s_inode_readahead_blks);
2221 
2222 	if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2223 		       (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2224 		SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2225 	if (nodefs || sbi->s_max_dir_size_kb)
2226 		SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2227 	if (test_opt(sb, DATA_ERR_ABORT))
2228 		SEQ_OPTS_PUTS("data_err=abort");
2229 	if (DUMMY_ENCRYPTION_ENABLED(sbi))
2230 		SEQ_OPTS_PUTS("test_dummy_encryption");
2231 
2232 	ext4_show_quota_options(seq, sb);
2233 	return 0;
2234 }
2235 
ext4_show_options(struct seq_file * seq,struct dentry * root)2236 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2237 {
2238 	return _ext4_show_options(seq, root->d_sb, 0);
2239 }
2240 
ext4_seq_options_show(struct seq_file * seq,void * offset)2241 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2242 {
2243 	struct super_block *sb = seq->private;
2244 	int rc;
2245 
2246 	seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2247 	rc = _ext4_show_options(seq, sb, 1);
2248 	seq_puts(seq, "\n");
2249 	return rc;
2250 }
2251 
ext4_setup_super(struct super_block * sb,struct ext4_super_block * es,int read_only)2252 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2253 			    int read_only)
2254 {
2255 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2256 	int err = 0;
2257 
2258 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2259 		ext4_msg(sb, KERN_ERR, "revision level too high, "
2260 			 "forcing read-only mode");
2261 		err = -EROFS;
2262 		goto done;
2263 	}
2264 	if (read_only)
2265 		goto done;
2266 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
2267 		ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2268 			 "running e2fsck is recommended");
2269 	else if (sbi->s_mount_state & EXT4_ERROR_FS)
2270 		ext4_msg(sb, KERN_WARNING,
2271 			 "warning: mounting fs with errors, "
2272 			 "running e2fsck is recommended");
2273 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2274 		 le16_to_cpu(es->s_mnt_count) >=
2275 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2276 		ext4_msg(sb, KERN_WARNING,
2277 			 "warning: maximal mount count reached, "
2278 			 "running e2fsck is recommended");
2279 	else if (le32_to_cpu(es->s_checkinterval) &&
2280 		 (ext4_get_tstamp(es, s_lastcheck) +
2281 		  le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2282 		ext4_msg(sb, KERN_WARNING,
2283 			 "warning: checktime reached, "
2284 			 "running e2fsck is recommended");
2285 	if (!sbi->s_journal)
2286 		es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2287 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2288 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2289 	le16_add_cpu(&es->s_mnt_count, 1);
2290 	ext4_update_tstamp(es, s_mtime);
2291 	if (sbi->s_journal)
2292 		ext4_set_feature_journal_needs_recovery(sb);
2293 
2294 	err = ext4_commit_super(sb, 1);
2295 done:
2296 	if (test_opt(sb, DEBUG))
2297 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2298 				"bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2299 			sb->s_blocksize,
2300 			sbi->s_groups_count,
2301 			EXT4_BLOCKS_PER_GROUP(sb),
2302 			EXT4_INODES_PER_GROUP(sb),
2303 			sbi->s_mount_opt, sbi->s_mount_opt2);
2304 
2305 	cleancache_init_fs(sb);
2306 	return err;
2307 }
2308 
ext4_alloc_flex_bg_array(struct super_block * sb,ext4_group_t ngroup)2309 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2310 {
2311 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2312 	struct flex_groups **old_groups, **new_groups;
2313 	int size, i, j;
2314 
2315 	if (!sbi->s_log_groups_per_flex)
2316 		return 0;
2317 
2318 	size = ext4_flex_group(sbi, ngroup - 1) + 1;
2319 	if (size <= sbi->s_flex_groups_allocated)
2320 		return 0;
2321 
2322 	new_groups = kvzalloc(roundup_pow_of_two(size *
2323 			      sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2324 	if (!new_groups) {
2325 		ext4_msg(sb, KERN_ERR,
2326 			 "not enough memory for %d flex group pointers", size);
2327 		return -ENOMEM;
2328 	}
2329 	for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2330 		new_groups[i] = kvzalloc(roundup_pow_of_two(
2331 					 sizeof(struct flex_groups)),
2332 					 GFP_KERNEL);
2333 		if (!new_groups[i]) {
2334 			for (j = sbi->s_flex_groups_allocated; j < i; j++)
2335 				kvfree(new_groups[j]);
2336 			kvfree(new_groups);
2337 			ext4_msg(sb, KERN_ERR,
2338 				 "not enough memory for %d flex groups", size);
2339 			return -ENOMEM;
2340 		}
2341 	}
2342 	rcu_read_lock();
2343 	old_groups = rcu_dereference(sbi->s_flex_groups);
2344 	if (old_groups)
2345 		memcpy(new_groups, old_groups,
2346 		       (sbi->s_flex_groups_allocated *
2347 			sizeof(struct flex_groups *)));
2348 	rcu_read_unlock();
2349 	rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2350 	sbi->s_flex_groups_allocated = size;
2351 	if (old_groups)
2352 		ext4_kvfree_array_rcu(old_groups);
2353 	return 0;
2354 }
2355 
ext4_fill_flex_info(struct super_block * sb)2356 static int ext4_fill_flex_info(struct super_block *sb)
2357 {
2358 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2359 	struct ext4_group_desc *gdp = NULL;
2360 	struct flex_groups *fg;
2361 	ext4_group_t flex_group;
2362 	int i, err;
2363 
2364 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2365 	if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2366 		sbi->s_log_groups_per_flex = 0;
2367 		return 1;
2368 	}
2369 
2370 	err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2371 	if (err)
2372 		goto failed;
2373 
2374 	for (i = 0; i < sbi->s_groups_count; i++) {
2375 		gdp = ext4_get_group_desc(sb, i, NULL);
2376 
2377 		flex_group = ext4_flex_group(sbi, i);
2378 		fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2379 		atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2380 		atomic64_add(ext4_free_group_clusters(sb, gdp),
2381 			     &fg->free_clusters);
2382 		atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2383 	}
2384 
2385 	return 1;
2386 failed:
2387 	return 0;
2388 }
2389 
ext4_group_desc_csum(struct super_block * sb,__u32 block_group,struct ext4_group_desc * gdp)2390 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2391 				   struct ext4_group_desc *gdp)
2392 {
2393 	int offset = offsetof(struct ext4_group_desc, bg_checksum);
2394 	__u16 crc = 0;
2395 	__le32 le_group = cpu_to_le32(block_group);
2396 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2397 
2398 	if (ext4_has_metadata_csum(sbi->s_sb)) {
2399 		/* Use new metadata_csum algorithm */
2400 		__u32 csum32;
2401 		__u16 dummy_csum = 0;
2402 
2403 		csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2404 				     sizeof(le_group));
2405 		csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2406 		csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2407 				     sizeof(dummy_csum));
2408 		offset += sizeof(dummy_csum);
2409 		if (offset < sbi->s_desc_size)
2410 			csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2411 					     sbi->s_desc_size - offset);
2412 
2413 		crc = csum32 & 0xFFFF;
2414 		goto out;
2415 	}
2416 
2417 	/* old crc16 code */
2418 	if (!ext4_has_feature_gdt_csum(sb))
2419 		return 0;
2420 
2421 	crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2422 	crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2423 	crc = crc16(crc, (__u8 *)gdp, offset);
2424 	offset += sizeof(gdp->bg_checksum); /* skip checksum */
2425 	/* for checksum of struct ext4_group_desc do the rest...*/
2426 	if (ext4_has_feature_64bit(sb) && offset < sbi->s_desc_size)
2427 		crc = crc16(crc, (__u8 *)gdp + offset,
2428 			    sbi->s_desc_size - offset);
2429 
2430 out:
2431 	return cpu_to_le16(crc);
2432 }
2433 
ext4_group_desc_csum_verify(struct super_block * sb,__u32 block_group,struct ext4_group_desc * gdp)2434 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2435 				struct ext4_group_desc *gdp)
2436 {
2437 	if (ext4_has_group_desc_csum(sb) &&
2438 	    (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2439 		return 0;
2440 
2441 	return 1;
2442 }
2443 
ext4_group_desc_csum_set(struct super_block * sb,__u32 block_group,struct ext4_group_desc * gdp)2444 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2445 			      struct ext4_group_desc *gdp)
2446 {
2447 	if (!ext4_has_group_desc_csum(sb))
2448 		return;
2449 	gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2450 }
2451 
2452 /* Called at mount-time, super-block is locked */
ext4_check_descriptors(struct super_block * sb,ext4_fsblk_t sb_block,ext4_group_t * first_not_zeroed)2453 static int ext4_check_descriptors(struct super_block *sb,
2454 				  ext4_fsblk_t sb_block,
2455 				  ext4_group_t *first_not_zeroed)
2456 {
2457 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2458 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2459 	ext4_fsblk_t last_block;
2460 	ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2461 	ext4_fsblk_t block_bitmap;
2462 	ext4_fsblk_t inode_bitmap;
2463 	ext4_fsblk_t inode_table;
2464 	int flexbg_flag = 0;
2465 	ext4_group_t i, grp = sbi->s_groups_count;
2466 
2467 	if (ext4_has_feature_flex_bg(sb))
2468 		flexbg_flag = 1;
2469 
2470 	ext4_debug("Checking group descriptors");
2471 
2472 	for (i = 0; i < sbi->s_groups_count; i++) {
2473 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2474 
2475 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
2476 			last_block = ext4_blocks_count(sbi->s_es) - 1;
2477 		else
2478 			last_block = first_block +
2479 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
2480 
2481 		if ((grp == sbi->s_groups_count) &&
2482 		   !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2483 			grp = i;
2484 
2485 		block_bitmap = ext4_block_bitmap(sb, gdp);
2486 		if (block_bitmap == sb_block) {
2487 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2488 				 "Block bitmap for group %u overlaps "
2489 				 "superblock", i);
2490 			if (!sb_rdonly(sb))
2491 				return 0;
2492 		}
2493 		if (block_bitmap >= sb_block + 1 &&
2494 		    block_bitmap <= last_bg_block) {
2495 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2496 				 "Block bitmap for group %u overlaps "
2497 				 "block group descriptors", i);
2498 			if (!sb_rdonly(sb))
2499 				return 0;
2500 		}
2501 		if (block_bitmap < first_block || block_bitmap > last_block) {
2502 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2503 			       "Block bitmap for group %u not in group "
2504 			       "(block %llu)!", i, block_bitmap);
2505 			return 0;
2506 		}
2507 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
2508 		if (inode_bitmap == sb_block) {
2509 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2510 				 "Inode bitmap for group %u overlaps "
2511 				 "superblock", i);
2512 			if (!sb_rdonly(sb))
2513 				return 0;
2514 		}
2515 		if (inode_bitmap >= sb_block + 1 &&
2516 		    inode_bitmap <= last_bg_block) {
2517 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2518 				 "Inode bitmap for group %u overlaps "
2519 				 "block group descriptors", i);
2520 			if (!sb_rdonly(sb))
2521 				return 0;
2522 		}
2523 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
2524 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2525 			       "Inode bitmap for group %u not in group "
2526 			       "(block %llu)!", i, inode_bitmap);
2527 			return 0;
2528 		}
2529 		inode_table = ext4_inode_table(sb, gdp);
2530 		if (inode_table == sb_block) {
2531 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2532 				 "Inode table for group %u overlaps "
2533 				 "superblock", i);
2534 			if (!sb_rdonly(sb))
2535 				return 0;
2536 		}
2537 		if (inode_table >= sb_block + 1 &&
2538 		    inode_table <= last_bg_block) {
2539 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2540 				 "Inode table for group %u overlaps "
2541 				 "block group descriptors", i);
2542 			if (!sb_rdonly(sb))
2543 				return 0;
2544 		}
2545 		if (inode_table < first_block ||
2546 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
2547 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2548 			       "Inode table for group %u not in group "
2549 			       "(block %llu)!", i, inode_table);
2550 			return 0;
2551 		}
2552 		ext4_lock_group(sb, i);
2553 		if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2554 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2555 				 "Checksum for group %u failed (%u!=%u)",
2556 				 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2557 				     gdp)), le16_to_cpu(gdp->bg_checksum));
2558 			if (!sb_rdonly(sb)) {
2559 				ext4_unlock_group(sb, i);
2560 				return 0;
2561 			}
2562 		}
2563 		ext4_unlock_group(sb, i);
2564 		if (!flexbg_flag)
2565 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
2566 	}
2567 	if (NULL != first_not_zeroed)
2568 		*first_not_zeroed = grp;
2569 	return 1;
2570 }
2571 
2572 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2573  * the superblock) which were deleted from all directories, but held open by
2574  * a process at the time of a crash.  We walk the list and try to delete these
2575  * inodes at recovery time (only with a read-write filesystem).
2576  *
2577  * In order to keep the orphan inode chain consistent during traversal (in
2578  * case of crash during recovery), we link each inode into the superblock
2579  * orphan list_head and handle it the same way as an inode deletion during
2580  * normal operation (which journals the operations for us).
2581  *
2582  * We only do an iget() and an iput() on each inode, which is very safe if we
2583  * accidentally point at an in-use or already deleted inode.  The worst that
2584  * can happen in this case is that we get a "bit already cleared" message from
2585  * ext4_free_inode().  The only reason we would point at a wrong inode is if
2586  * e2fsck was run on this filesystem, and it must have already done the orphan
2587  * inode cleanup for us, so we can safely abort without any further action.
2588  */
ext4_orphan_cleanup(struct super_block * sb,struct ext4_super_block * es)2589 static void ext4_orphan_cleanup(struct super_block *sb,
2590 				struct ext4_super_block *es)
2591 {
2592 	unsigned int s_flags = sb->s_flags;
2593 	int ret, nr_orphans = 0, nr_truncates = 0;
2594 #ifdef CONFIG_QUOTA
2595 	int quota_update = 0;
2596 	int i;
2597 #endif
2598 	if (!es->s_last_orphan) {
2599 		jbd_debug(4, "no orphan inodes to clean up\n");
2600 		return;
2601 	}
2602 
2603 	if (bdev_read_only(sb->s_bdev)) {
2604 		ext4_msg(sb, KERN_ERR, "write access "
2605 			"unavailable, skipping orphan cleanup");
2606 		return;
2607 	}
2608 
2609 	/* Check if feature set would not allow a r/w mount */
2610 	if (!ext4_feature_set_ok(sb, 0)) {
2611 		ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2612 			 "unknown ROCOMPAT features");
2613 		return;
2614 	}
2615 
2616 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2617 		/* don't clear list on RO mount w/ errors */
2618 		if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2619 			ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2620 				  "clearing orphan list.\n");
2621 			es->s_last_orphan = 0;
2622 		}
2623 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2624 		return;
2625 	}
2626 
2627 	if (s_flags & SB_RDONLY) {
2628 		ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2629 		sb->s_flags &= ~SB_RDONLY;
2630 	}
2631 #ifdef CONFIG_QUOTA
2632 	/*
2633 	 * Turn on quotas which were not enabled for read-only mounts if
2634 	 * filesystem has quota feature, so that they are updated correctly.
2635 	 */
2636 	if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
2637 		int ret = ext4_enable_quotas(sb);
2638 
2639 		if (!ret)
2640 			quota_update = 1;
2641 		else
2642 			ext4_msg(sb, KERN_ERR,
2643 				"Cannot turn on quotas: error %d", ret);
2644 	}
2645 
2646 	/* Turn on journaled quotas used for old sytle */
2647 	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2648 		if (EXT4_SB(sb)->s_qf_names[i]) {
2649 			int ret = ext4_quota_on_mount(sb, i);
2650 
2651 			if (!ret)
2652 				quota_update = 1;
2653 			else
2654 				ext4_msg(sb, KERN_ERR,
2655 					"Cannot turn on journaled "
2656 					"quota: type %d: error %d", i, ret);
2657 		}
2658 	}
2659 #endif
2660 
2661 	while (es->s_last_orphan) {
2662 		struct inode *inode;
2663 
2664 		/*
2665 		 * We may have encountered an error during cleanup; if
2666 		 * so, skip the rest.
2667 		 */
2668 		if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2669 			jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2670 			es->s_last_orphan = 0;
2671 			break;
2672 		}
2673 
2674 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2675 		if (IS_ERR(inode)) {
2676 			es->s_last_orphan = 0;
2677 			break;
2678 		}
2679 
2680 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2681 		dquot_initialize(inode);
2682 		if (inode->i_nlink) {
2683 			if (test_opt(sb, DEBUG))
2684 				ext4_msg(sb, KERN_DEBUG,
2685 					"%s: truncating inode %lu to %lld bytes",
2686 					__func__, inode->i_ino, inode->i_size);
2687 			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2688 				  inode->i_ino, inode->i_size);
2689 			inode_lock(inode);
2690 			truncate_inode_pages(inode->i_mapping, inode->i_size);
2691 			ret = ext4_truncate(inode);
2692 			if (ret) {
2693 				/*
2694 				 * We need to clean up the in-core orphan list
2695 				 * manually if ext4_truncate() failed to get a
2696 				 * transaction handle.
2697 				 */
2698 				ext4_orphan_del(NULL, inode);
2699 				ext4_std_error(inode->i_sb, ret);
2700 			}
2701 			inode_unlock(inode);
2702 			nr_truncates++;
2703 		} else {
2704 			if (test_opt(sb, DEBUG))
2705 				ext4_msg(sb, KERN_DEBUG,
2706 					"%s: deleting unreferenced inode %lu",
2707 					__func__, inode->i_ino);
2708 			jbd_debug(2, "deleting unreferenced inode %lu\n",
2709 				  inode->i_ino);
2710 			nr_orphans++;
2711 		}
2712 		iput(inode);  /* The delete magic happens here! */
2713 	}
2714 
2715 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2716 
2717 	if (nr_orphans)
2718 		ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2719 		       PLURAL(nr_orphans));
2720 	if (nr_truncates)
2721 		ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2722 		       PLURAL(nr_truncates));
2723 #ifdef CONFIG_QUOTA
2724 	/* Turn off quotas if they were enabled for orphan cleanup */
2725 	if (quota_update) {
2726 		for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2727 			if (sb_dqopt(sb)->files[i])
2728 				dquot_quota_off(sb, i);
2729 		}
2730 	}
2731 #endif
2732 	sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2733 }
2734 
2735 /*
2736  * Maximal extent format file size.
2737  * Resulting logical blkno at s_maxbytes must fit in our on-disk
2738  * extent format containers, within a sector_t, and within i_blocks
2739  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2740  * so that won't be a limiting factor.
2741  *
2742  * However there is other limiting factor. We do store extents in the form
2743  * of starting block and length, hence the resulting length of the extent
2744  * covering maximum file size must fit into on-disk format containers as
2745  * well. Given that length is always by 1 unit bigger than max unit (because
2746  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2747  *
2748  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2749  */
ext4_max_size(int blkbits,int has_huge_files)2750 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2751 {
2752 	loff_t res;
2753 	loff_t upper_limit = MAX_LFS_FILESIZE;
2754 
2755 	/* small i_blocks in vfs inode? */
2756 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2757 		/*
2758 		 * CONFIG_LBDAF is not enabled implies the inode
2759 		 * i_block represent total blocks in 512 bytes
2760 		 * 32 == size of vfs inode i_blocks * 8
2761 		 */
2762 		upper_limit = (1LL << 32) - 1;
2763 
2764 		/* total blocks in file system block size */
2765 		upper_limit >>= (blkbits - 9);
2766 		upper_limit <<= blkbits;
2767 	}
2768 
2769 	/*
2770 	 * 32-bit extent-start container, ee_block. We lower the maxbytes
2771 	 * by one fs block, so ee_len can cover the extent of maximum file
2772 	 * size
2773 	 */
2774 	res = (1LL << 32) - 1;
2775 	res <<= blkbits;
2776 
2777 	/* Sanity check against vm- & vfs- imposed limits */
2778 	if (res > upper_limit)
2779 		res = upper_limit;
2780 
2781 	return res;
2782 }
2783 
2784 /*
2785  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2786  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2787  * We need to be 1 filesystem block less than the 2^48 sector limit.
2788  */
ext4_max_bitmap_size(int bits,int has_huge_files)2789 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2790 {
2791 	loff_t res = EXT4_NDIR_BLOCKS;
2792 	int meta_blocks;
2793 	loff_t upper_limit;
2794 	/* This is calculated to be the largest file size for a dense, block
2795 	 * mapped file such that the file's total number of 512-byte sectors,
2796 	 * including data and all indirect blocks, does not exceed (2^48 - 1).
2797 	 *
2798 	 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2799 	 * number of 512-byte sectors of the file.
2800 	 */
2801 
2802 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2803 		/*
2804 		 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2805 		 * the inode i_block field represents total file blocks in
2806 		 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2807 		 */
2808 		upper_limit = (1LL << 32) - 1;
2809 
2810 		/* total blocks in file system block size */
2811 		upper_limit >>= (bits - 9);
2812 
2813 	} else {
2814 		/*
2815 		 * We use 48 bit ext4_inode i_blocks
2816 		 * With EXT4_HUGE_FILE_FL set the i_blocks
2817 		 * represent total number of blocks in
2818 		 * file system block size
2819 		 */
2820 		upper_limit = (1LL << 48) - 1;
2821 
2822 	}
2823 
2824 	/* indirect blocks */
2825 	meta_blocks = 1;
2826 	/* double indirect blocks */
2827 	meta_blocks += 1 + (1LL << (bits-2));
2828 	/* tripple indirect blocks */
2829 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2830 
2831 	upper_limit -= meta_blocks;
2832 	upper_limit <<= bits;
2833 
2834 	res += 1LL << (bits-2);
2835 	res += 1LL << (2*(bits-2));
2836 	res += 1LL << (3*(bits-2));
2837 	res <<= bits;
2838 	if (res > upper_limit)
2839 		res = upper_limit;
2840 
2841 	if (res > MAX_LFS_FILESIZE)
2842 		res = MAX_LFS_FILESIZE;
2843 
2844 	return res;
2845 }
2846 
descriptor_loc(struct super_block * sb,ext4_fsblk_t logical_sb_block,int nr)2847 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2848 				   ext4_fsblk_t logical_sb_block, int nr)
2849 {
2850 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2851 	ext4_group_t bg, first_meta_bg;
2852 	int has_super = 0;
2853 
2854 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2855 
2856 	if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2857 		return logical_sb_block + nr + 1;
2858 	bg = sbi->s_desc_per_block * nr;
2859 	if (ext4_bg_has_super(sb, bg))
2860 		has_super = 1;
2861 
2862 	/*
2863 	 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2864 	 * block 2, not 1.  If s_first_data_block == 0 (bigalloc is enabled
2865 	 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2866 	 * compensate.
2867 	 */
2868 	if (sb->s_blocksize == 1024 && nr == 0 &&
2869 	    le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
2870 		has_super++;
2871 
2872 	return (has_super + ext4_group_first_block_no(sb, bg));
2873 }
2874 
2875 /**
2876  * ext4_get_stripe_size: Get the stripe size.
2877  * @sbi: In memory super block info
2878  *
2879  * If we have specified it via mount option, then
2880  * use the mount option value. If the value specified at mount time is
2881  * greater than the blocks per group use the super block value.
2882  * If the super block value is greater than blocks per group return 0.
2883  * Allocator needs it be less than blocks per group.
2884  *
2885  */
ext4_get_stripe_size(struct ext4_sb_info * sbi)2886 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2887 {
2888 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2889 	unsigned long stripe_width =
2890 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2891 	int ret;
2892 
2893 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2894 		ret = sbi->s_stripe;
2895 	else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2896 		ret = stripe_width;
2897 	else if (stride && stride <= sbi->s_blocks_per_group)
2898 		ret = stride;
2899 	else
2900 		ret = 0;
2901 
2902 	/*
2903 	 * If the stripe width is 1, this makes no sense and
2904 	 * we set it to 0 to turn off stripe handling code.
2905 	 */
2906 	if (ret <= 1)
2907 		ret = 0;
2908 
2909 	return ret;
2910 }
2911 
2912 /*
2913  * Check whether this filesystem can be mounted based on
2914  * the features present and the RDONLY/RDWR mount requested.
2915  * Returns 1 if this filesystem can be mounted as requested,
2916  * 0 if it cannot be.
2917  */
ext4_feature_set_ok(struct super_block * sb,int readonly)2918 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2919 {
2920 	if (ext4_has_unknown_ext4_incompat_features(sb)) {
2921 		ext4_msg(sb, KERN_ERR,
2922 			"Couldn't mount because of "
2923 			"unsupported optional features (%x)",
2924 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2925 			~EXT4_FEATURE_INCOMPAT_SUPP));
2926 		return 0;
2927 	}
2928 
2929 	if (readonly)
2930 		return 1;
2931 
2932 	if (ext4_has_feature_readonly(sb)) {
2933 		ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2934 		sb->s_flags |= SB_RDONLY;
2935 		return 1;
2936 	}
2937 
2938 	/* Check that feature set is OK for a read-write mount */
2939 	if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2940 		ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2941 			 "unsupported optional features (%x)",
2942 			 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2943 				~EXT4_FEATURE_RO_COMPAT_SUPP));
2944 		return 0;
2945 	}
2946 	/*
2947 	 * Large file size enabled file system can only be mounted
2948 	 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2949 	 */
2950 	if (ext4_has_feature_huge_file(sb)) {
2951 		if (sizeof(blkcnt_t) < sizeof(u64)) {
2952 			ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2953 				 "cannot be mounted RDWR without "
2954 				 "CONFIG_LBDAF");
2955 			return 0;
2956 		}
2957 	}
2958 	if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2959 		ext4_msg(sb, KERN_ERR,
2960 			 "Can't support bigalloc feature without "
2961 			 "extents feature\n");
2962 		return 0;
2963 	}
2964 
2965 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
2966 	if (!readonly && (ext4_has_feature_quota(sb) ||
2967 			  ext4_has_feature_project(sb))) {
2968 		ext4_msg(sb, KERN_ERR,
2969 			 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
2970 		return 0;
2971 	}
2972 #endif  /* CONFIG_QUOTA */
2973 	return 1;
2974 }
2975 
2976 /*
2977  * This function is called once a day if we have errors logged
2978  * on the file system
2979  */
print_daily_error_info(struct timer_list * t)2980 static void print_daily_error_info(struct timer_list *t)
2981 {
2982 	struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
2983 	struct super_block *sb = sbi->s_sb;
2984 	struct ext4_super_block *es = sbi->s_es;
2985 
2986 	if (es->s_error_count)
2987 		/* fsck newer than v1.41.13 is needed to clean this condition. */
2988 		ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2989 			 le32_to_cpu(es->s_error_count));
2990 	if (es->s_first_error_time) {
2991 		printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
2992 		       sb->s_id,
2993 		       ext4_get_tstamp(es, s_first_error_time),
2994 		       (int) sizeof(es->s_first_error_func),
2995 		       es->s_first_error_func,
2996 		       le32_to_cpu(es->s_first_error_line));
2997 		if (es->s_first_error_ino)
2998 			printk(KERN_CONT ": inode %u",
2999 			       le32_to_cpu(es->s_first_error_ino));
3000 		if (es->s_first_error_block)
3001 			printk(KERN_CONT ": block %llu", (unsigned long long)
3002 			       le64_to_cpu(es->s_first_error_block));
3003 		printk(KERN_CONT "\n");
3004 	}
3005 	if (es->s_last_error_time) {
3006 		printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3007 		       sb->s_id,
3008 		       ext4_get_tstamp(es, s_last_error_time),
3009 		       (int) sizeof(es->s_last_error_func),
3010 		       es->s_last_error_func,
3011 		       le32_to_cpu(es->s_last_error_line));
3012 		if (es->s_last_error_ino)
3013 			printk(KERN_CONT ": inode %u",
3014 			       le32_to_cpu(es->s_last_error_ino));
3015 		if (es->s_last_error_block)
3016 			printk(KERN_CONT ": block %llu", (unsigned long long)
3017 			       le64_to_cpu(es->s_last_error_block));
3018 		printk(KERN_CONT "\n");
3019 	}
3020 	mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
3021 }
3022 
3023 /* Find next suitable group and run ext4_init_inode_table */
ext4_run_li_request(struct ext4_li_request * elr)3024 static int ext4_run_li_request(struct ext4_li_request *elr)
3025 {
3026 	struct ext4_group_desc *gdp = NULL;
3027 	ext4_group_t group, ngroups;
3028 	struct super_block *sb;
3029 	int ret = 0;
3030 	u64 start_time;
3031 
3032 	sb = elr->lr_super;
3033 	ngroups = EXT4_SB(sb)->s_groups_count;
3034 
3035 	for (group = elr->lr_next_group; group < ngroups; group++) {
3036 		gdp = ext4_get_group_desc(sb, group, NULL);
3037 		if (!gdp) {
3038 			ret = 1;
3039 			break;
3040 		}
3041 
3042 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3043 			break;
3044 	}
3045 
3046 	if (group >= ngroups)
3047 		ret = 1;
3048 
3049 	if (!ret) {
3050 		start_time = ktime_get_real_ns();
3051 		ret = ext4_init_inode_table(sb, group,
3052 					    elr->lr_timeout ? 0 : 1);
3053 		if (elr->lr_timeout == 0) {
3054 			elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3055 				  elr->lr_sbi->s_li_wait_mult);
3056 		}
3057 		elr->lr_next_sched = jiffies + elr->lr_timeout;
3058 		elr->lr_next_group = group + 1;
3059 	}
3060 	return ret;
3061 }
3062 
3063 /*
3064  * Remove lr_request from the list_request and free the
3065  * request structure. Should be called with li_list_mtx held
3066  */
ext4_remove_li_request(struct ext4_li_request * elr)3067 static void ext4_remove_li_request(struct ext4_li_request *elr)
3068 {
3069 	struct ext4_sb_info *sbi;
3070 
3071 	if (!elr)
3072 		return;
3073 
3074 	sbi = elr->lr_sbi;
3075 
3076 	list_del(&elr->lr_request);
3077 	sbi->s_li_request = NULL;
3078 	kfree(elr);
3079 }
3080 
ext4_unregister_li_request(struct super_block * sb)3081 static void ext4_unregister_li_request(struct super_block *sb)
3082 {
3083 	mutex_lock(&ext4_li_mtx);
3084 	if (!ext4_li_info) {
3085 		mutex_unlock(&ext4_li_mtx);
3086 		return;
3087 	}
3088 
3089 	mutex_lock(&ext4_li_info->li_list_mtx);
3090 	ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3091 	mutex_unlock(&ext4_li_info->li_list_mtx);
3092 	mutex_unlock(&ext4_li_mtx);
3093 }
3094 
3095 static struct task_struct *ext4_lazyinit_task;
3096 
3097 /*
3098  * This is the function where ext4lazyinit thread lives. It walks
3099  * through the request list searching for next scheduled filesystem.
3100  * When such a fs is found, run the lazy initialization request
3101  * (ext4_rn_li_request) and keep track of the time spend in this
3102  * function. Based on that time we compute next schedule time of
3103  * the request. When walking through the list is complete, compute
3104  * next waking time and put itself into sleep.
3105  */
ext4_lazyinit_thread(void * arg)3106 static int ext4_lazyinit_thread(void *arg)
3107 {
3108 	struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3109 	struct list_head *pos, *n;
3110 	struct ext4_li_request *elr;
3111 	unsigned long next_wakeup, cur;
3112 
3113 	BUG_ON(NULL == eli);
3114 	set_freezable();
3115 
3116 cont_thread:
3117 	while (true) {
3118 		next_wakeup = MAX_JIFFY_OFFSET;
3119 
3120 		mutex_lock(&eli->li_list_mtx);
3121 		if (list_empty(&eli->li_request_list)) {
3122 			mutex_unlock(&eli->li_list_mtx);
3123 			goto exit_thread;
3124 		}
3125 		list_for_each_safe(pos, n, &eli->li_request_list) {
3126 			int err = 0;
3127 			int progress = 0;
3128 			elr = list_entry(pos, struct ext4_li_request,
3129 					 lr_request);
3130 
3131 			if (time_before(jiffies, elr->lr_next_sched)) {
3132 				if (time_before(elr->lr_next_sched, next_wakeup))
3133 					next_wakeup = elr->lr_next_sched;
3134 				continue;
3135 			}
3136 			if (down_read_trylock(&elr->lr_super->s_umount)) {
3137 				if (sb_start_write_trylock(elr->lr_super)) {
3138 					progress = 1;
3139 					/*
3140 					 * We hold sb->s_umount, sb can not
3141 					 * be removed from the list, it is
3142 					 * now safe to drop li_list_mtx
3143 					 */
3144 					mutex_unlock(&eli->li_list_mtx);
3145 					err = ext4_run_li_request(elr);
3146 					sb_end_write(elr->lr_super);
3147 					mutex_lock(&eli->li_list_mtx);
3148 					n = pos->next;
3149 				}
3150 				up_read((&elr->lr_super->s_umount));
3151 			}
3152 			/* error, remove the lazy_init job */
3153 			if (err) {
3154 				ext4_remove_li_request(elr);
3155 				continue;
3156 			}
3157 			if (!progress) {
3158 				elr->lr_next_sched = jiffies +
3159 					(prandom_u32()
3160 					 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3161 			}
3162 			if (time_before(elr->lr_next_sched, next_wakeup))
3163 				next_wakeup = elr->lr_next_sched;
3164 		}
3165 		mutex_unlock(&eli->li_list_mtx);
3166 
3167 		try_to_freeze();
3168 
3169 		cur = jiffies;
3170 		if ((time_after_eq(cur, next_wakeup)) ||
3171 		    (MAX_JIFFY_OFFSET == next_wakeup)) {
3172 			cond_resched();
3173 			continue;
3174 		}
3175 
3176 		schedule_timeout_interruptible(next_wakeup - cur);
3177 
3178 		if (kthread_should_stop()) {
3179 			ext4_clear_request_list();
3180 			goto exit_thread;
3181 		}
3182 	}
3183 
3184 exit_thread:
3185 	/*
3186 	 * It looks like the request list is empty, but we need
3187 	 * to check it under the li_list_mtx lock, to prevent any
3188 	 * additions into it, and of course we should lock ext4_li_mtx
3189 	 * to atomically free the list and ext4_li_info, because at
3190 	 * this point another ext4 filesystem could be registering
3191 	 * new one.
3192 	 */
3193 	mutex_lock(&ext4_li_mtx);
3194 	mutex_lock(&eli->li_list_mtx);
3195 	if (!list_empty(&eli->li_request_list)) {
3196 		mutex_unlock(&eli->li_list_mtx);
3197 		mutex_unlock(&ext4_li_mtx);
3198 		goto cont_thread;
3199 	}
3200 	mutex_unlock(&eli->li_list_mtx);
3201 	kfree(ext4_li_info);
3202 	ext4_li_info = NULL;
3203 	mutex_unlock(&ext4_li_mtx);
3204 
3205 	return 0;
3206 }
3207 
ext4_clear_request_list(void)3208 static void ext4_clear_request_list(void)
3209 {
3210 	struct list_head *pos, *n;
3211 	struct ext4_li_request *elr;
3212 
3213 	mutex_lock(&ext4_li_info->li_list_mtx);
3214 	list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3215 		elr = list_entry(pos, struct ext4_li_request,
3216 				 lr_request);
3217 		ext4_remove_li_request(elr);
3218 	}
3219 	mutex_unlock(&ext4_li_info->li_list_mtx);
3220 }
3221 
ext4_run_lazyinit_thread(void)3222 static int ext4_run_lazyinit_thread(void)
3223 {
3224 	ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3225 					 ext4_li_info, "ext4lazyinit");
3226 	if (IS_ERR(ext4_lazyinit_task)) {
3227 		int err = PTR_ERR(ext4_lazyinit_task);
3228 		ext4_clear_request_list();
3229 		kfree(ext4_li_info);
3230 		ext4_li_info = NULL;
3231 		printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3232 				 "initialization thread\n",
3233 				 err);
3234 		return err;
3235 	}
3236 	ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3237 	return 0;
3238 }
3239 
3240 /*
3241  * Check whether it make sense to run itable init. thread or not.
3242  * If there is at least one uninitialized inode table, return
3243  * corresponding group number, else the loop goes through all
3244  * groups and return total number of groups.
3245  */
ext4_has_uninit_itable(struct super_block * sb)3246 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3247 {
3248 	ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3249 	struct ext4_group_desc *gdp = NULL;
3250 
3251 	if (!ext4_has_group_desc_csum(sb))
3252 		return ngroups;
3253 
3254 	for (group = 0; group < ngroups; group++) {
3255 		gdp = ext4_get_group_desc(sb, group, NULL);
3256 		if (!gdp)
3257 			continue;
3258 
3259 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3260 			break;
3261 	}
3262 
3263 	return group;
3264 }
3265 
ext4_li_info_new(void)3266 static int ext4_li_info_new(void)
3267 {
3268 	struct ext4_lazy_init *eli = NULL;
3269 
3270 	eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3271 	if (!eli)
3272 		return -ENOMEM;
3273 
3274 	INIT_LIST_HEAD(&eli->li_request_list);
3275 	mutex_init(&eli->li_list_mtx);
3276 
3277 	eli->li_state |= EXT4_LAZYINIT_QUIT;
3278 
3279 	ext4_li_info = eli;
3280 
3281 	return 0;
3282 }
3283 
ext4_li_request_new(struct super_block * sb,ext4_group_t start)3284 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3285 					    ext4_group_t start)
3286 {
3287 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3288 	struct ext4_li_request *elr;
3289 
3290 	elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3291 	if (!elr)
3292 		return NULL;
3293 
3294 	elr->lr_super = sb;
3295 	elr->lr_sbi = sbi;
3296 	elr->lr_next_group = start;
3297 
3298 	/*
3299 	 * Randomize first schedule time of the request to
3300 	 * spread the inode table initialization requests
3301 	 * better.
3302 	 */
3303 	elr->lr_next_sched = jiffies + (prandom_u32() %
3304 				(EXT4_DEF_LI_MAX_START_DELAY * HZ));
3305 	return elr;
3306 }
3307 
ext4_register_li_request(struct super_block * sb,ext4_group_t first_not_zeroed)3308 int ext4_register_li_request(struct super_block *sb,
3309 			     ext4_group_t first_not_zeroed)
3310 {
3311 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3312 	struct ext4_li_request *elr = NULL;
3313 	ext4_group_t ngroups = sbi->s_groups_count;
3314 	int ret = 0;
3315 
3316 	mutex_lock(&ext4_li_mtx);
3317 	if (sbi->s_li_request != NULL) {
3318 		/*
3319 		 * Reset timeout so it can be computed again, because
3320 		 * s_li_wait_mult might have changed.
3321 		 */
3322 		sbi->s_li_request->lr_timeout = 0;
3323 		goto out;
3324 	}
3325 
3326 	if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3327 	    !test_opt(sb, INIT_INODE_TABLE))
3328 		goto out;
3329 
3330 	elr = ext4_li_request_new(sb, first_not_zeroed);
3331 	if (!elr) {
3332 		ret = -ENOMEM;
3333 		goto out;
3334 	}
3335 
3336 	if (NULL == ext4_li_info) {
3337 		ret = ext4_li_info_new();
3338 		if (ret)
3339 			goto out;
3340 	}
3341 
3342 	mutex_lock(&ext4_li_info->li_list_mtx);
3343 	list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3344 	mutex_unlock(&ext4_li_info->li_list_mtx);
3345 
3346 	sbi->s_li_request = elr;
3347 	/*
3348 	 * set elr to NULL here since it has been inserted to
3349 	 * the request_list and the removal and free of it is
3350 	 * handled by ext4_clear_request_list from now on.
3351 	 */
3352 	elr = NULL;
3353 
3354 	if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3355 		ret = ext4_run_lazyinit_thread();
3356 		if (ret)
3357 			goto out;
3358 	}
3359 out:
3360 	mutex_unlock(&ext4_li_mtx);
3361 	if (ret)
3362 		kfree(elr);
3363 	return ret;
3364 }
3365 
3366 /*
3367  * We do not need to lock anything since this is called on
3368  * module unload.
3369  */
ext4_destroy_lazyinit_thread(void)3370 static void ext4_destroy_lazyinit_thread(void)
3371 {
3372 	/*
3373 	 * If thread exited earlier
3374 	 * there's nothing to be done.
3375 	 */
3376 	if (!ext4_li_info || !ext4_lazyinit_task)
3377 		return;
3378 
3379 	kthread_stop(ext4_lazyinit_task);
3380 }
3381 
set_journal_csum_feature_set(struct super_block * sb)3382 static int set_journal_csum_feature_set(struct super_block *sb)
3383 {
3384 	int ret = 1;
3385 	int compat, incompat;
3386 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3387 
3388 	if (ext4_has_metadata_csum(sb)) {
3389 		/* journal checksum v3 */
3390 		compat = 0;
3391 		incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3392 	} else {
3393 		/* journal checksum v1 */
3394 		compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3395 		incompat = 0;
3396 	}
3397 
3398 	jbd2_journal_clear_features(sbi->s_journal,
3399 			JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3400 			JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3401 			JBD2_FEATURE_INCOMPAT_CSUM_V2);
3402 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3403 		ret = jbd2_journal_set_features(sbi->s_journal,
3404 				compat, 0,
3405 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3406 				incompat);
3407 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3408 		ret = jbd2_journal_set_features(sbi->s_journal,
3409 				compat, 0,
3410 				incompat);
3411 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3412 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3413 	} else {
3414 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3415 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3416 	}
3417 
3418 	return ret;
3419 }
3420 
3421 /*
3422  * Note: calculating the overhead so we can be compatible with
3423  * historical BSD practice is quite difficult in the face of
3424  * clusters/bigalloc.  This is because multiple metadata blocks from
3425  * different block group can end up in the same allocation cluster.
3426  * Calculating the exact overhead in the face of clustered allocation
3427  * requires either O(all block bitmaps) in memory or O(number of block
3428  * groups**2) in time.  We will still calculate the superblock for
3429  * older file systems --- and if we come across with a bigalloc file
3430  * system with zero in s_overhead_clusters the estimate will be close to
3431  * correct especially for very large cluster sizes --- but for newer
3432  * file systems, it's better to calculate this figure once at mkfs
3433  * time, and store it in the superblock.  If the superblock value is
3434  * present (even for non-bigalloc file systems), we will use it.
3435  */
count_overhead(struct super_block * sb,ext4_group_t grp,char * buf)3436 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3437 			  char *buf)
3438 {
3439 	struct ext4_sb_info	*sbi = EXT4_SB(sb);
3440 	struct ext4_group_desc	*gdp;
3441 	ext4_fsblk_t		first_block, last_block, b;
3442 	ext4_group_t		i, ngroups = ext4_get_groups_count(sb);
3443 	int			s, j, count = 0;
3444 	int			has_super = ext4_bg_has_super(sb, grp);
3445 
3446 	if (!ext4_has_feature_bigalloc(sb))
3447 		return (has_super + ext4_bg_num_gdb(sb, grp) +
3448 			(has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
3449 			sbi->s_itb_per_group + 2);
3450 
3451 	first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3452 		(grp * EXT4_BLOCKS_PER_GROUP(sb));
3453 	last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3454 	for (i = 0; i < ngroups; i++) {
3455 		gdp = ext4_get_group_desc(sb, i, NULL);
3456 		b = ext4_block_bitmap(sb, gdp);
3457 		if (b >= first_block && b <= last_block) {
3458 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3459 			count++;
3460 		}
3461 		b = ext4_inode_bitmap(sb, gdp);
3462 		if (b >= first_block && b <= last_block) {
3463 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3464 			count++;
3465 		}
3466 		b = ext4_inode_table(sb, gdp);
3467 		if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3468 			for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3469 				int c = EXT4_B2C(sbi, b - first_block);
3470 				ext4_set_bit(c, buf);
3471 				count++;
3472 			}
3473 		if (i != grp)
3474 			continue;
3475 		s = 0;
3476 		if (ext4_bg_has_super(sb, grp)) {
3477 			ext4_set_bit(s++, buf);
3478 			count++;
3479 		}
3480 		j = ext4_bg_num_gdb(sb, grp);
3481 		if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3482 			ext4_error(sb, "Invalid number of block group "
3483 				   "descriptor blocks: %d", j);
3484 			j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3485 		}
3486 		count += j;
3487 		for (; j > 0; j--)
3488 			ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3489 	}
3490 	if (!count)
3491 		return 0;
3492 	return EXT4_CLUSTERS_PER_GROUP(sb) -
3493 		ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3494 }
3495 
3496 /*
3497  * Compute the overhead and stash it in sbi->s_overhead
3498  */
ext4_calculate_overhead(struct super_block * sb)3499 int ext4_calculate_overhead(struct super_block *sb)
3500 {
3501 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3502 	struct ext4_super_block *es = sbi->s_es;
3503 	struct inode *j_inode;
3504 	unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3505 	ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3506 	ext4_fsblk_t overhead = 0;
3507 	char *buf = (char *) get_zeroed_page(GFP_NOFS);
3508 
3509 	if (!buf)
3510 		return -ENOMEM;
3511 
3512 	/*
3513 	 * Compute the overhead (FS structures).  This is constant
3514 	 * for a given filesystem unless the number of block groups
3515 	 * changes so we cache the previous value until it does.
3516 	 */
3517 
3518 	/*
3519 	 * All of the blocks before first_data_block are overhead
3520 	 */
3521 	overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3522 
3523 	/*
3524 	 * Add the overhead found in each block group
3525 	 */
3526 	for (i = 0; i < ngroups; i++) {
3527 		int blks;
3528 
3529 		blks = count_overhead(sb, i, buf);
3530 		overhead += blks;
3531 		if (blks)
3532 			memset(buf, 0, PAGE_SIZE);
3533 		cond_resched();
3534 	}
3535 
3536 	/*
3537 	 * Add the internal journal blocks whether the journal has been
3538 	 * loaded or not
3539 	 */
3540 	if (sbi->s_journal && !sbi->s_journal_bdev)
3541 		overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3542 	else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3543 		/* j_inum for internal journal is non-zero */
3544 		j_inode = ext4_get_journal_inode(sb, j_inum);
3545 		if (j_inode) {
3546 			j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3547 			overhead += EXT4_NUM_B2C(sbi, j_blocks);
3548 			iput(j_inode);
3549 		} else {
3550 			ext4_msg(sb, KERN_ERR, "can't get journal size");
3551 		}
3552 	}
3553 	sbi->s_overhead = overhead;
3554 	smp_wmb();
3555 	free_page((unsigned long) buf);
3556 	return 0;
3557 }
3558 
ext4_set_resv_clusters(struct super_block * sb)3559 static void ext4_set_resv_clusters(struct super_block *sb)
3560 {
3561 	ext4_fsblk_t resv_clusters;
3562 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3563 
3564 	/*
3565 	 * There's no need to reserve anything when we aren't using extents.
3566 	 * The space estimates are exact, there are no unwritten extents,
3567 	 * hole punching doesn't need new metadata... This is needed especially
3568 	 * to keep ext2/3 backward compatibility.
3569 	 */
3570 	if (!ext4_has_feature_extents(sb))
3571 		return;
3572 	/*
3573 	 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3574 	 * This should cover the situations where we can not afford to run
3575 	 * out of space like for example punch hole, or converting
3576 	 * unwritten extents in delalloc path. In most cases such
3577 	 * allocation would require 1, or 2 blocks, higher numbers are
3578 	 * very rare.
3579 	 */
3580 	resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3581 			 sbi->s_cluster_bits);
3582 
3583 	do_div(resv_clusters, 50);
3584 	resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3585 
3586 	atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3587 }
3588 
ext4_fill_super(struct super_block * sb,void * data,int silent)3589 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3590 {
3591 	struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3592 	char *orig_data = kstrdup(data, GFP_KERNEL);
3593 	struct buffer_head *bh, **group_desc;
3594 	struct ext4_super_block *es = NULL;
3595 	struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3596 	struct flex_groups **flex_groups;
3597 	ext4_fsblk_t block;
3598 	ext4_fsblk_t sb_block = get_sb_block(&data);
3599 	ext4_fsblk_t logical_sb_block;
3600 	unsigned long offset = 0;
3601 	unsigned long journal_devnum = 0;
3602 	unsigned long def_mount_opts;
3603 	struct inode *root;
3604 	const char *descr;
3605 	int ret = -ENOMEM;
3606 	int blocksize, clustersize;
3607 	unsigned int db_count;
3608 	unsigned int i;
3609 	int needs_recovery, has_huge_files, has_bigalloc;
3610 	__u64 blocks_count;
3611 	int err = 0;
3612 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3613 	ext4_group_t first_not_zeroed;
3614 
3615 	if ((data && !orig_data) || !sbi)
3616 		goto out_free_base;
3617 
3618 	sbi->s_daxdev = dax_dev;
3619 	sbi->s_blockgroup_lock =
3620 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3621 	if (!sbi->s_blockgroup_lock)
3622 		goto out_free_base;
3623 
3624 	sb->s_fs_info = sbi;
3625 	sbi->s_sb = sb;
3626 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3627 	sbi->s_sb_block = sb_block;
3628 	if (sb->s_bdev->bd_part)
3629 		sbi->s_sectors_written_start =
3630 			part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
3631 
3632 	/* Cleanup superblock name */
3633 	strreplace(sb->s_id, '/', '!');
3634 
3635 	/* -EINVAL is default */
3636 	ret = -EINVAL;
3637 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3638 	if (!blocksize) {
3639 		ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3640 		goto out_fail;
3641 	}
3642 
3643 	/*
3644 	 * The ext4 superblock will not be buffer aligned for other than 1kB
3645 	 * block sizes.  We need to calculate the offset from buffer start.
3646 	 */
3647 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3648 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3649 		offset = do_div(logical_sb_block, blocksize);
3650 	} else {
3651 		logical_sb_block = sb_block;
3652 	}
3653 
3654 	if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3655 		ext4_msg(sb, KERN_ERR, "unable to read superblock");
3656 		goto out_fail;
3657 	}
3658 	/*
3659 	 * Note: s_es must be initialized as soon as possible because
3660 	 *       some ext4 macro-instructions depend on its value
3661 	 */
3662 	es = (struct ext4_super_block *) (bh->b_data + offset);
3663 	sbi->s_es = es;
3664 	sb->s_magic = le16_to_cpu(es->s_magic);
3665 	if (sb->s_magic != EXT4_SUPER_MAGIC)
3666 		goto cantfind_ext4;
3667 	sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3668 
3669 	/* Warn if metadata_csum and gdt_csum are both set. */
3670 	if (ext4_has_feature_metadata_csum(sb) &&
3671 	    ext4_has_feature_gdt_csum(sb))
3672 		ext4_warning(sb, "metadata_csum and uninit_bg are "
3673 			     "redundant flags; please run fsck.");
3674 
3675 	/* Check for a known checksum algorithm */
3676 	if (!ext4_verify_csum_type(sb, es)) {
3677 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3678 			 "unknown checksum algorithm.");
3679 		silent = 1;
3680 		goto cantfind_ext4;
3681 	}
3682 
3683 	/* Load the checksum driver */
3684 	sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3685 	if (IS_ERR(sbi->s_chksum_driver)) {
3686 		ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3687 		ret = PTR_ERR(sbi->s_chksum_driver);
3688 		sbi->s_chksum_driver = NULL;
3689 		goto failed_mount;
3690 	}
3691 
3692 	/* Check superblock checksum */
3693 	if (!ext4_superblock_csum_verify(sb, es)) {
3694 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3695 			 "invalid superblock checksum.  Run e2fsck?");
3696 		silent = 1;
3697 		ret = -EFSBADCRC;
3698 		goto cantfind_ext4;
3699 	}
3700 
3701 	/* Precompute checksum seed for all metadata */
3702 	if (ext4_has_feature_csum_seed(sb))
3703 		sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3704 	else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3705 		sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3706 					       sizeof(es->s_uuid));
3707 
3708 	/* Set defaults before we parse the mount options */
3709 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3710 	set_opt(sb, INIT_INODE_TABLE);
3711 	if (def_mount_opts & EXT4_DEFM_DEBUG)
3712 		set_opt(sb, DEBUG);
3713 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3714 		set_opt(sb, GRPID);
3715 	if (def_mount_opts & EXT4_DEFM_UID16)
3716 		set_opt(sb, NO_UID32);
3717 	/* xattr user namespace & acls are now defaulted on */
3718 	set_opt(sb, XATTR_USER);
3719 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3720 	set_opt(sb, POSIX_ACL);
3721 #endif
3722 	/* don't forget to enable journal_csum when metadata_csum is enabled. */
3723 	if (ext4_has_metadata_csum(sb))
3724 		set_opt(sb, JOURNAL_CHECKSUM);
3725 
3726 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3727 		set_opt(sb, JOURNAL_DATA);
3728 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3729 		set_opt(sb, ORDERED_DATA);
3730 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3731 		set_opt(sb, WRITEBACK_DATA);
3732 
3733 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3734 		set_opt(sb, ERRORS_PANIC);
3735 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3736 		set_opt(sb, ERRORS_CONT);
3737 	else
3738 		set_opt(sb, ERRORS_RO);
3739 	/* block_validity enabled by default; disable with noblock_validity */
3740 	set_opt(sb, BLOCK_VALIDITY);
3741 	if (def_mount_opts & EXT4_DEFM_DISCARD)
3742 		set_opt(sb, DISCARD);
3743 
3744 	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3745 	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3746 	sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3747 	sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3748 	sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3749 
3750 	if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3751 		set_opt(sb, BARRIER);
3752 
3753 	/*
3754 	 * enable delayed allocation by default
3755 	 * Use -o nodelalloc to turn it off
3756 	 */
3757 	if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3758 	    ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3759 		set_opt(sb, DELALLOC);
3760 
3761 	/*
3762 	 * set default s_li_wait_mult for lazyinit, for the case there is
3763 	 * no mount option specified.
3764 	 */
3765 	sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3766 
3767 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3768 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3769 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
3770 		ext4_msg(sb, KERN_ERR,
3771 		       "Unsupported filesystem blocksize %d (%d log_block_size)",
3772 			 blocksize, le32_to_cpu(es->s_log_block_size));
3773 		goto failed_mount;
3774 	}
3775 
3776 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3777 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3778 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3779 	} else {
3780 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3781 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3782 		if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3783 			ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3784 				 sbi->s_first_ino);
3785 			goto failed_mount;
3786 		}
3787 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3788 		    (!is_power_of_2(sbi->s_inode_size)) ||
3789 		    (sbi->s_inode_size > blocksize)) {
3790 			ext4_msg(sb, KERN_ERR,
3791 			       "unsupported inode size: %d",
3792 			       sbi->s_inode_size);
3793 			ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
3794 			goto failed_mount;
3795 		}
3796 		/*
3797 		 * i_atime_extra is the last extra field available for
3798 		 * [acm]times in struct ext4_inode. Checking for that
3799 		 * field should suffice to ensure we have extra space
3800 		 * for all three.
3801 		 */
3802 		if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
3803 			sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
3804 			sb->s_time_gran = 1;
3805 		} else {
3806 			sb->s_time_gran = NSEC_PER_SEC;
3807 		}
3808 	}
3809 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3810 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3811 			EXT4_GOOD_OLD_INODE_SIZE;
3812 		if (ext4_has_feature_extra_isize(sb)) {
3813 			unsigned v, max = (sbi->s_inode_size -
3814 					   EXT4_GOOD_OLD_INODE_SIZE);
3815 
3816 			v = le16_to_cpu(es->s_want_extra_isize);
3817 			if (v > max) {
3818 				ext4_msg(sb, KERN_ERR,
3819 					 "bad s_want_extra_isize: %d", v);
3820 				goto failed_mount;
3821 			}
3822 			if (sbi->s_want_extra_isize < v)
3823 				sbi->s_want_extra_isize = v;
3824 
3825 			v = le16_to_cpu(es->s_min_extra_isize);
3826 			if (v > max) {
3827 				ext4_msg(sb, KERN_ERR,
3828 					 "bad s_min_extra_isize: %d", v);
3829 				goto failed_mount;
3830 			}
3831 			if (sbi->s_want_extra_isize < v)
3832 				sbi->s_want_extra_isize = v;
3833 		}
3834 	}
3835 
3836 	if (sbi->s_es->s_mount_opts[0]) {
3837 		char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3838 					      sizeof(sbi->s_es->s_mount_opts),
3839 					      GFP_KERNEL);
3840 		if (!s_mount_opts)
3841 			goto failed_mount;
3842 		if (!parse_options(s_mount_opts, sb, &journal_devnum,
3843 				   &journal_ioprio, 0)) {
3844 			ext4_msg(sb, KERN_WARNING,
3845 				 "failed to parse options in superblock: %s",
3846 				 s_mount_opts);
3847 		}
3848 		kfree(s_mount_opts);
3849 	}
3850 	sbi->s_def_mount_opt = sbi->s_mount_opt;
3851 	if (!parse_options((char *) data, sb, &journal_devnum,
3852 			   &journal_ioprio, 0))
3853 		goto failed_mount;
3854 
3855 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3856 		printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3857 			    "with data=journal disables delayed "
3858 			    "allocation and O_DIRECT support!\n");
3859 		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3860 			ext4_msg(sb, KERN_ERR, "can't mount with "
3861 				 "both data=journal and delalloc");
3862 			goto failed_mount;
3863 		}
3864 		if (test_opt(sb, DIOREAD_NOLOCK)) {
3865 			ext4_msg(sb, KERN_ERR, "can't mount with "
3866 				 "both data=journal and dioread_nolock");
3867 			goto failed_mount;
3868 		}
3869 		if (test_opt(sb, DAX)) {
3870 			ext4_msg(sb, KERN_ERR, "can't mount with "
3871 				 "both data=journal and dax");
3872 			goto failed_mount;
3873 		}
3874 		if (ext4_has_feature_encrypt(sb)) {
3875 			ext4_msg(sb, KERN_WARNING,
3876 				 "encrypted files will use data=ordered "
3877 				 "instead of data journaling mode");
3878 		}
3879 		if (test_opt(sb, DELALLOC))
3880 			clear_opt(sb, DELALLOC);
3881 	} else {
3882 		sb->s_iflags |= SB_I_CGROUPWB;
3883 	}
3884 
3885 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3886 		(test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
3887 
3888 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3889 	    (ext4_has_compat_features(sb) ||
3890 	     ext4_has_ro_compat_features(sb) ||
3891 	     ext4_has_incompat_features(sb)))
3892 		ext4_msg(sb, KERN_WARNING,
3893 		       "feature flags set on rev 0 fs, "
3894 		       "running e2fsck is recommended");
3895 
3896 	if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3897 		set_opt2(sb, HURD_COMPAT);
3898 		if (ext4_has_feature_64bit(sb)) {
3899 			ext4_msg(sb, KERN_ERR,
3900 				 "The Hurd can't support 64-bit file systems");
3901 			goto failed_mount;
3902 		}
3903 
3904 		/*
3905 		 * ea_inode feature uses l_i_version field which is not
3906 		 * available in HURD_COMPAT mode.
3907 		 */
3908 		if (ext4_has_feature_ea_inode(sb)) {
3909 			ext4_msg(sb, KERN_ERR,
3910 				 "ea_inode feature is not supported for Hurd");
3911 			goto failed_mount;
3912 		}
3913 	}
3914 
3915 	if (IS_EXT2_SB(sb)) {
3916 		if (ext2_feature_set_ok(sb))
3917 			ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3918 				 "using the ext4 subsystem");
3919 		else {
3920 			/*
3921 			 * If we're probing be silent, if this looks like
3922 			 * it's actually an ext[34] filesystem.
3923 			 */
3924 			if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3925 				goto failed_mount;
3926 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3927 				 "to feature incompatibilities");
3928 			goto failed_mount;
3929 		}
3930 	}
3931 
3932 	if (IS_EXT3_SB(sb)) {
3933 		if (ext3_feature_set_ok(sb))
3934 			ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3935 				 "using the ext4 subsystem");
3936 		else {
3937 			/*
3938 			 * If we're probing be silent, if this looks like
3939 			 * it's actually an ext4 filesystem.
3940 			 */
3941 			if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3942 				goto failed_mount;
3943 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3944 				 "to feature incompatibilities");
3945 			goto failed_mount;
3946 		}
3947 	}
3948 
3949 	/*
3950 	 * Check feature flags regardless of the revision level, since we
3951 	 * previously didn't change the revision level when setting the flags,
3952 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
3953 	 */
3954 	if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
3955 		goto failed_mount;
3956 
3957 	if (le32_to_cpu(es->s_log_block_size) >
3958 	    (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3959 		ext4_msg(sb, KERN_ERR,
3960 			 "Invalid log block size: %u",
3961 			 le32_to_cpu(es->s_log_block_size));
3962 		goto failed_mount;
3963 	}
3964 	if (le32_to_cpu(es->s_log_cluster_size) >
3965 	    (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3966 		ext4_msg(sb, KERN_ERR,
3967 			 "Invalid log cluster size: %u",
3968 			 le32_to_cpu(es->s_log_cluster_size));
3969 		goto failed_mount;
3970 	}
3971 
3972 	if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3973 		ext4_msg(sb, KERN_ERR,
3974 			 "Number of reserved GDT blocks insanely large: %d",
3975 			 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3976 		goto failed_mount;
3977 	}
3978 
3979 	if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3980 		if (ext4_has_feature_inline_data(sb)) {
3981 			ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
3982 					" that may contain inline data");
3983 			sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
3984 		}
3985 		if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
3986 			ext4_msg(sb, KERN_ERR,
3987 				"DAX unsupported by block device. Turning off DAX.");
3988 			sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
3989 		}
3990 	}
3991 
3992 	if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3993 		ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3994 			 es->s_encryption_level);
3995 		goto failed_mount;
3996 	}
3997 
3998 	if (sb->s_blocksize != blocksize) {
3999 		/* Validate the filesystem blocksize */
4000 		if (!sb_set_blocksize(sb, blocksize)) {
4001 			ext4_msg(sb, KERN_ERR, "bad block size %d",
4002 					blocksize);
4003 			goto failed_mount;
4004 		}
4005 
4006 		brelse(bh);
4007 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4008 		offset = do_div(logical_sb_block, blocksize);
4009 		bh = sb_bread_unmovable(sb, logical_sb_block);
4010 		if (!bh) {
4011 			ext4_msg(sb, KERN_ERR,
4012 			       "Can't read superblock on 2nd try");
4013 			goto failed_mount;
4014 		}
4015 		es = (struct ext4_super_block *)(bh->b_data + offset);
4016 		sbi->s_es = es;
4017 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4018 			ext4_msg(sb, KERN_ERR,
4019 			       "Magic mismatch, very weird!");
4020 			goto failed_mount;
4021 		}
4022 	}
4023 
4024 	has_huge_files = ext4_has_feature_huge_file(sb);
4025 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4026 						      has_huge_files);
4027 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4028 
4029 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4030 	if (ext4_has_feature_64bit(sb)) {
4031 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4032 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4033 		    !is_power_of_2(sbi->s_desc_size)) {
4034 			ext4_msg(sb, KERN_ERR,
4035 			       "unsupported descriptor size %lu",
4036 			       sbi->s_desc_size);
4037 			goto failed_mount;
4038 		}
4039 	} else
4040 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4041 
4042 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4043 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4044 
4045 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4046 	if (sbi->s_inodes_per_block == 0)
4047 		goto cantfind_ext4;
4048 	if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4049 	    sbi->s_inodes_per_group > blocksize * 8) {
4050 		ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4051 			 sbi->s_inodes_per_group);
4052 		goto failed_mount;
4053 	}
4054 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
4055 					sbi->s_inodes_per_block;
4056 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4057 	sbi->s_sbh = bh;
4058 	sbi->s_mount_state = le16_to_cpu(es->s_state);
4059 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4060 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4061 
4062 	for (i = 0; i < 4; i++)
4063 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4064 	sbi->s_def_hash_version = es->s_def_hash_version;
4065 	if (ext4_has_feature_dir_index(sb)) {
4066 		i = le32_to_cpu(es->s_flags);
4067 		if (i & EXT2_FLAGS_UNSIGNED_HASH)
4068 			sbi->s_hash_unsigned = 3;
4069 		else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4070 #ifdef __CHAR_UNSIGNED__
4071 			if (!sb_rdonly(sb))
4072 				es->s_flags |=
4073 					cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4074 			sbi->s_hash_unsigned = 3;
4075 #else
4076 			if (!sb_rdonly(sb))
4077 				es->s_flags |=
4078 					cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4079 #endif
4080 		}
4081 	}
4082 
4083 	/* Handle clustersize */
4084 	clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4085 	has_bigalloc = ext4_has_feature_bigalloc(sb);
4086 	if (has_bigalloc) {
4087 		if (clustersize < blocksize) {
4088 			ext4_msg(sb, KERN_ERR,
4089 				 "cluster size (%d) smaller than "
4090 				 "block size (%d)", clustersize, blocksize);
4091 			goto failed_mount;
4092 		}
4093 		sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4094 			le32_to_cpu(es->s_log_block_size);
4095 		sbi->s_clusters_per_group =
4096 			le32_to_cpu(es->s_clusters_per_group);
4097 		if (sbi->s_clusters_per_group > blocksize * 8) {
4098 			ext4_msg(sb, KERN_ERR,
4099 				 "#clusters per group too big: %lu",
4100 				 sbi->s_clusters_per_group);
4101 			goto failed_mount;
4102 		}
4103 		if (sbi->s_blocks_per_group !=
4104 		    (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4105 			ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4106 				 "clusters per group (%lu) inconsistent",
4107 				 sbi->s_blocks_per_group,
4108 				 sbi->s_clusters_per_group);
4109 			goto failed_mount;
4110 		}
4111 	} else {
4112 		if (clustersize != blocksize) {
4113 			ext4_msg(sb, KERN_ERR,
4114 				 "fragment/cluster size (%d) != "
4115 				 "block size (%d)", clustersize, blocksize);
4116 			goto failed_mount;
4117 		}
4118 		if (sbi->s_blocks_per_group > blocksize * 8) {
4119 			ext4_msg(sb, KERN_ERR,
4120 				 "#blocks per group too big: %lu",
4121 				 sbi->s_blocks_per_group);
4122 			goto failed_mount;
4123 		}
4124 		sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4125 		sbi->s_cluster_bits = 0;
4126 	}
4127 	sbi->s_cluster_ratio = clustersize / blocksize;
4128 
4129 	/* Do we have standard group size of clustersize * 8 blocks ? */
4130 	if (sbi->s_blocks_per_group == clustersize << 3)
4131 		set_opt2(sb, STD_GROUP_SIZE);
4132 
4133 	/*
4134 	 * Test whether we have more sectors than will fit in sector_t,
4135 	 * and whether the max offset is addressable by the page cache.
4136 	 */
4137 	err = generic_check_addressable(sb->s_blocksize_bits,
4138 					ext4_blocks_count(es));
4139 	if (err) {
4140 		ext4_msg(sb, KERN_ERR, "filesystem"
4141 			 " too large to mount safely on this system");
4142 		if (sizeof(sector_t) < 8)
4143 			ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
4144 		goto failed_mount;
4145 	}
4146 
4147 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4148 		goto cantfind_ext4;
4149 
4150 	/* check blocks count against device size */
4151 	blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4152 	if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4153 		ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4154 		       "exceeds size of device (%llu blocks)",
4155 		       ext4_blocks_count(es), blocks_count);
4156 		goto failed_mount;
4157 	}
4158 
4159 	/*
4160 	 * It makes no sense for the first data block to be beyond the end
4161 	 * of the filesystem.
4162 	 */
4163 	if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4164 		ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4165 			 "block %u is beyond end of filesystem (%llu)",
4166 			 le32_to_cpu(es->s_first_data_block),
4167 			 ext4_blocks_count(es));
4168 		goto failed_mount;
4169 	}
4170 	if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4171 	    (sbi->s_cluster_ratio == 1)) {
4172 		ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4173 			 "block is 0 with a 1k block and cluster size");
4174 		goto failed_mount;
4175 	}
4176 
4177 	blocks_count = (ext4_blocks_count(es) -
4178 			le32_to_cpu(es->s_first_data_block) +
4179 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
4180 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4181 	if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4182 		ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4183 		       "(block count %llu, first data block %u, "
4184 		       "blocks per group %lu)", blocks_count,
4185 		       ext4_blocks_count(es),
4186 		       le32_to_cpu(es->s_first_data_block),
4187 		       EXT4_BLOCKS_PER_GROUP(sb));
4188 		goto failed_mount;
4189 	}
4190 	sbi->s_groups_count = blocks_count;
4191 	sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4192 			(EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4193 	if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4194 	    le32_to_cpu(es->s_inodes_count)) {
4195 		ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4196 			 le32_to_cpu(es->s_inodes_count),
4197 			 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4198 		ret = -EINVAL;
4199 		goto failed_mount;
4200 	}
4201 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4202 		   EXT4_DESC_PER_BLOCK(sb);
4203 	if (ext4_has_feature_meta_bg(sb)) {
4204 		if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4205 			ext4_msg(sb, KERN_WARNING,
4206 				 "first meta block group too large: %u "
4207 				 "(group descriptor block count %u)",
4208 				 le32_to_cpu(es->s_first_meta_bg), db_count);
4209 			goto failed_mount;
4210 		}
4211 	}
4212 	rcu_assign_pointer(sbi->s_group_desc,
4213 			   kvmalloc_array(db_count,
4214 					  sizeof(struct buffer_head *),
4215 					  GFP_KERNEL));
4216 	if (sbi->s_group_desc == NULL) {
4217 		ext4_msg(sb, KERN_ERR, "not enough memory");
4218 		ret = -ENOMEM;
4219 		goto failed_mount;
4220 	}
4221 
4222 	bgl_lock_init(sbi->s_blockgroup_lock);
4223 
4224 	/* Pre-read the descriptors into the buffer cache */
4225 	for (i = 0; i < db_count; i++) {
4226 		block = descriptor_loc(sb, logical_sb_block, i);
4227 		sb_breadahead_unmovable(sb, block);
4228 	}
4229 
4230 	for (i = 0; i < db_count; i++) {
4231 		struct buffer_head *bh;
4232 
4233 		block = descriptor_loc(sb, logical_sb_block, i);
4234 		bh = sb_bread_unmovable(sb, block);
4235 		if (!bh) {
4236 			ext4_msg(sb, KERN_ERR,
4237 			       "can't read group descriptor %d", i);
4238 			db_count = i;
4239 			goto failed_mount2;
4240 		}
4241 		rcu_read_lock();
4242 		rcu_dereference(sbi->s_group_desc)[i] = bh;
4243 		rcu_read_unlock();
4244 	}
4245 	sbi->s_gdb_count = db_count;
4246 	if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4247 		ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4248 		ret = -EFSCORRUPTED;
4249 		goto failed_mount2;
4250 	}
4251 
4252 	timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4253 
4254 	/* Register extent status tree shrinker */
4255 	if (ext4_es_register_shrinker(sbi))
4256 		goto failed_mount3;
4257 
4258 	sbi->s_stripe = ext4_get_stripe_size(sbi);
4259 	sbi->s_extent_max_zeroout_kb = 32;
4260 
4261 	/*
4262 	 * set up enough so that it can read an inode
4263 	 */
4264 	sb->s_op = &ext4_sops;
4265 	sb->s_export_op = &ext4_export_ops;
4266 	sb->s_xattr = ext4_xattr_handlers;
4267 #ifdef CONFIG_EXT4_FS_ENCRYPTION
4268 	sb->s_cop = &ext4_cryptops;
4269 #endif
4270 #ifdef CONFIG_QUOTA
4271 	sb->dq_op = &ext4_quota_operations;
4272 	if (ext4_has_feature_quota(sb))
4273 		sb->s_qcop = &dquot_quotactl_sysfile_ops;
4274 	else
4275 		sb->s_qcop = &ext4_qctl_operations;
4276 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4277 #endif
4278 	memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4279 
4280 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4281 	mutex_init(&sbi->s_orphan_lock);
4282 
4283 	sb->s_root = NULL;
4284 
4285 	needs_recovery = (es->s_last_orphan != 0 ||
4286 			  ext4_has_feature_journal_needs_recovery(sb));
4287 
4288 	if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4289 		if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4290 			goto failed_mount3a;
4291 
4292 	/*
4293 	 * The first inode we look at is the journal inode.  Don't try
4294 	 * root first: it may be modified in the journal!
4295 	 */
4296 	if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4297 		err = ext4_load_journal(sb, es, journal_devnum);
4298 		if (err)
4299 			goto failed_mount3a;
4300 	} else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4301 		   ext4_has_feature_journal_needs_recovery(sb)) {
4302 		ext4_msg(sb, KERN_ERR, "required journal recovery "
4303 		       "suppressed and not mounted read-only");
4304 		goto failed_mount3a;
4305 	} else {
4306 		/* Nojournal mode, all journal mount options are illegal */
4307 		if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4308 			ext4_msg(sb, KERN_ERR, "can't mount with "
4309 				 "journal_async_commit, fs mounted w/o journal");
4310 			goto failed_mount3a;
4311 		}
4312 
4313 		if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4314 			ext4_msg(sb, KERN_ERR, "can't mount with "
4315 				 "journal_checksum, fs mounted w/o journal");
4316 			goto failed_mount3a;
4317 		}
4318 		if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4319 			ext4_msg(sb, KERN_ERR, "can't mount with "
4320 				 "commit=%lu, fs mounted w/o journal",
4321 				 sbi->s_commit_interval / HZ);
4322 			goto failed_mount3a;
4323 		}
4324 		if (EXT4_MOUNT_DATA_FLAGS &
4325 		    (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4326 			ext4_msg(sb, KERN_ERR, "can't mount with "
4327 				 "data=, fs mounted w/o journal");
4328 			goto failed_mount3a;
4329 		}
4330 		sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4331 		clear_opt(sb, JOURNAL_CHECKSUM);
4332 		clear_opt(sb, DATA_FLAGS);
4333 		sbi->s_journal = NULL;
4334 		needs_recovery = 0;
4335 		goto no_journal;
4336 	}
4337 
4338 	if (ext4_has_feature_64bit(sb) &&
4339 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4340 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
4341 		ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4342 		goto failed_mount_wq;
4343 	}
4344 
4345 	if (!set_journal_csum_feature_set(sb)) {
4346 		ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4347 			 "feature set");
4348 		goto failed_mount_wq;
4349 	}
4350 
4351 	/* We have now updated the journal if required, so we can
4352 	 * validate the data journaling mode. */
4353 	switch (test_opt(sb, DATA_FLAGS)) {
4354 	case 0:
4355 		/* No mode set, assume a default based on the journal
4356 		 * capabilities: ORDERED_DATA if the journal can
4357 		 * cope, else JOURNAL_DATA
4358 		 */
4359 		if (jbd2_journal_check_available_features
4360 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4361 			set_opt(sb, ORDERED_DATA);
4362 			sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4363 		} else {
4364 			set_opt(sb, JOURNAL_DATA);
4365 			sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4366 		}
4367 		break;
4368 
4369 	case EXT4_MOUNT_ORDERED_DATA:
4370 	case EXT4_MOUNT_WRITEBACK_DATA:
4371 		if (!jbd2_journal_check_available_features
4372 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4373 			ext4_msg(sb, KERN_ERR, "Journal does not support "
4374 			       "requested data journaling mode");
4375 			goto failed_mount_wq;
4376 		}
4377 	default:
4378 		break;
4379 	}
4380 
4381 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4382 	    test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4383 		ext4_msg(sb, KERN_ERR, "can't mount with "
4384 			"journal_async_commit in data=ordered mode");
4385 		goto failed_mount_wq;
4386 	}
4387 
4388 	set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4389 
4390 	sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4391 
4392 no_journal:
4393 	if (!test_opt(sb, NO_MBCACHE)) {
4394 		sbi->s_ea_block_cache = ext4_xattr_create_cache();
4395 		if (!sbi->s_ea_block_cache) {
4396 			ext4_msg(sb, KERN_ERR,
4397 				 "Failed to create ea_block_cache");
4398 			goto failed_mount_wq;
4399 		}
4400 
4401 		if (ext4_has_feature_ea_inode(sb)) {
4402 			sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4403 			if (!sbi->s_ea_inode_cache) {
4404 				ext4_msg(sb, KERN_ERR,
4405 					 "Failed to create ea_inode_cache");
4406 				goto failed_mount_wq;
4407 			}
4408 		}
4409 	}
4410 
4411 	if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4412 	    (blocksize != PAGE_SIZE)) {
4413 		ext4_msg(sb, KERN_ERR,
4414 			 "Unsupported blocksize for fs encryption");
4415 		goto failed_mount_wq;
4416 	}
4417 
4418 	if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4419 	    !ext4_has_feature_encrypt(sb)) {
4420 		ext4_set_feature_encrypt(sb);
4421 		ext4_commit_super(sb, 1);
4422 	}
4423 
4424 	/*
4425 	 * Get the # of file system overhead blocks from the
4426 	 * superblock if present.
4427 	 */
4428 	sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4429 	/* ignore the precalculated value if it is ridiculous */
4430 	if (sbi->s_overhead > ext4_blocks_count(es))
4431 		sbi->s_overhead = 0;
4432 	/*
4433 	 * If the bigalloc feature is not enabled recalculating the
4434 	 * overhead doesn't take long, so we might as well just redo
4435 	 * it to make sure we are using the correct value.
4436 	 */
4437 	if (!ext4_has_feature_bigalloc(sb))
4438 		sbi->s_overhead = 0;
4439 	if (sbi->s_overhead == 0) {
4440 		err = ext4_calculate_overhead(sb);
4441 		if (err)
4442 			goto failed_mount_wq;
4443 	}
4444 
4445 	/*
4446 	 * The maximum number of concurrent works can be high and
4447 	 * concurrency isn't really necessary.  Limit it to 1.
4448 	 */
4449 	EXT4_SB(sb)->rsv_conversion_wq =
4450 		alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4451 	if (!EXT4_SB(sb)->rsv_conversion_wq) {
4452 		printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4453 		ret = -ENOMEM;
4454 		goto failed_mount4;
4455 	}
4456 
4457 	/*
4458 	 * The jbd2_journal_load will have done any necessary log recovery,
4459 	 * so we can safely mount the rest of the filesystem now.
4460 	 */
4461 
4462 	root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4463 	if (IS_ERR(root)) {
4464 		ext4_msg(sb, KERN_ERR, "get root inode failed");
4465 		ret = PTR_ERR(root);
4466 		root = NULL;
4467 		goto failed_mount4;
4468 	}
4469 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4470 		ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4471 		iput(root);
4472 		goto failed_mount4;
4473 	}
4474 	sb->s_root = d_make_root(root);
4475 	if (!sb->s_root) {
4476 		ext4_msg(sb, KERN_ERR, "get root dentry failed");
4477 		ret = -ENOMEM;
4478 		goto failed_mount4;
4479 	}
4480 
4481 	ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4482 	if (ret == -EROFS) {
4483 		sb->s_flags |= SB_RDONLY;
4484 		ret = 0;
4485 	} else if (ret)
4486 		goto failed_mount4a;
4487 
4488 	ext4_set_resv_clusters(sb);
4489 
4490 	if (test_opt(sb, BLOCK_VALIDITY)) {
4491 		err = ext4_setup_system_zone(sb);
4492 		if (err) {
4493 			ext4_msg(sb, KERN_ERR, "failed to initialize system "
4494 				 "zone (%d)", err);
4495 			goto failed_mount4a;
4496 		}
4497 	}
4498 
4499 	ext4_ext_init(sb);
4500 	err = ext4_mb_init(sb);
4501 	if (err) {
4502 		ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4503 			 err);
4504 		goto failed_mount5;
4505 	}
4506 
4507 	block = ext4_count_free_clusters(sb);
4508 	ext4_free_blocks_count_set(sbi->s_es,
4509 				   EXT4_C2B(sbi, block));
4510 	ext4_superblock_csum_set(sb);
4511 	err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4512 				  GFP_KERNEL);
4513 	if (!err) {
4514 		unsigned long freei = ext4_count_free_inodes(sb);
4515 		sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4516 		ext4_superblock_csum_set(sb);
4517 		err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4518 					  GFP_KERNEL);
4519 	}
4520 	if (!err)
4521 		err = percpu_counter_init(&sbi->s_dirs_counter,
4522 					  ext4_count_dirs(sb), GFP_KERNEL);
4523 	if (!err)
4524 		err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4525 					  GFP_KERNEL);
4526 	if (!err)
4527 		err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
4528 
4529 	if (err) {
4530 		ext4_msg(sb, KERN_ERR, "insufficient memory");
4531 		goto failed_mount6;
4532 	}
4533 
4534 	if (ext4_has_feature_flex_bg(sb))
4535 		if (!ext4_fill_flex_info(sb)) {
4536 			ext4_msg(sb, KERN_ERR,
4537 			       "unable to initialize "
4538 			       "flex_bg meta info!");
4539 			ret = -ENOMEM;
4540 			goto failed_mount6;
4541 		}
4542 
4543 	err = ext4_register_li_request(sb, first_not_zeroed);
4544 	if (err)
4545 		goto failed_mount6;
4546 
4547 	err = ext4_register_sysfs(sb);
4548 	if (err)
4549 		goto failed_mount7;
4550 
4551 #ifdef CONFIG_QUOTA
4552 	/* Enable quota usage during mount. */
4553 	if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4554 		err = ext4_enable_quotas(sb);
4555 		if (err)
4556 			goto failed_mount8;
4557 	}
4558 #endif  /* CONFIG_QUOTA */
4559 
4560 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4561 	ext4_orphan_cleanup(sb, es);
4562 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4563 	if (needs_recovery) {
4564 		ext4_msg(sb, KERN_INFO, "recovery complete");
4565 		err = ext4_mark_recovery_complete(sb, es);
4566 		if (err)
4567 			goto failed_mount8;
4568 	}
4569 	if (EXT4_SB(sb)->s_journal) {
4570 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4571 			descr = " journalled data mode";
4572 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4573 			descr = " ordered data mode";
4574 		else
4575 			descr = " writeback data mode";
4576 	} else
4577 		descr = "out journal";
4578 
4579 	if (test_opt(sb, DISCARD)) {
4580 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
4581 		if (!blk_queue_discard(q))
4582 			ext4_msg(sb, KERN_WARNING,
4583 				 "mounting with \"discard\" option, but "
4584 				 "the device does not support discard");
4585 	}
4586 
4587 	if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4588 		ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4589 			 "Opts: %.*s%s%s", descr,
4590 			 (int) sizeof(sbi->s_es->s_mount_opts),
4591 			 sbi->s_es->s_mount_opts,
4592 			 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4593 
4594 	if (es->s_error_count)
4595 		mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4596 
4597 	/* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4598 	ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4599 	ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4600 	ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4601 
4602 	kfree(orig_data);
4603 	return 0;
4604 
4605 cantfind_ext4:
4606 	if (!silent)
4607 		ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4608 	goto failed_mount;
4609 
4610 failed_mount8:
4611 	ext4_unregister_sysfs(sb);
4612 	kobject_put(&sbi->s_kobj);
4613 failed_mount7:
4614 	ext4_unregister_li_request(sb);
4615 failed_mount6:
4616 	ext4_mb_release(sb);
4617 	rcu_read_lock();
4618 	flex_groups = rcu_dereference(sbi->s_flex_groups);
4619 	if (flex_groups) {
4620 		for (i = 0; i < sbi->s_flex_groups_allocated; i++)
4621 			kvfree(flex_groups[i]);
4622 		kvfree(flex_groups);
4623 	}
4624 	rcu_read_unlock();
4625 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
4626 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
4627 	percpu_counter_destroy(&sbi->s_dirs_counter);
4628 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4629 	percpu_free_rwsem(&sbi->s_writepages_rwsem);
4630 failed_mount5:
4631 	ext4_ext_release(sb);
4632 	ext4_release_system_zone(sb);
4633 failed_mount4a:
4634 	dput(sb->s_root);
4635 	sb->s_root = NULL;
4636 failed_mount4:
4637 	ext4_msg(sb, KERN_ERR, "mount failed");
4638 	if (EXT4_SB(sb)->rsv_conversion_wq)
4639 		destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4640 failed_mount_wq:
4641 	if (sbi->s_ea_inode_cache) {
4642 		ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4643 		sbi->s_ea_inode_cache = NULL;
4644 	}
4645 	if (sbi->s_ea_block_cache) {
4646 		ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4647 		sbi->s_ea_block_cache = NULL;
4648 	}
4649 	if (sbi->s_journal) {
4650 		jbd2_journal_destroy(sbi->s_journal);
4651 		sbi->s_journal = NULL;
4652 	}
4653 failed_mount3a:
4654 	ext4_es_unregister_shrinker(sbi);
4655 failed_mount3:
4656 	del_timer_sync(&sbi->s_err_report);
4657 	if (sbi->s_mmp_tsk)
4658 		kthread_stop(sbi->s_mmp_tsk);
4659 failed_mount2:
4660 	rcu_read_lock();
4661 	group_desc = rcu_dereference(sbi->s_group_desc);
4662 	for (i = 0; i < db_count; i++)
4663 		brelse(group_desc[i]);
4664 	kvfree(group_desc);
4665 	rcu_read_unlock();
4666 failed_mount:
4667 	if (sbi->s_chksum_driver)
4668 		crypto_free_shash(sbi->s_chksum_driver);
4669 #ifdef CONFIG_QUOTA
4670 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
4671 		kfree(sbi->s_qf_names[i]);
4672 #endif
4673 	ext4_blkdev_remove(sbi);
4674 	brelse(bh);
4675 out_fail:
4676 	invalidate_bdev(sb->s_bdev);
4677 	sb->s_fs_info = NULL;
4678 	kfree(sbi->s_blockgroup_lock);
4679 out_free_base:
4680 	kfree(sbi);
4681 	kfree(orig_data);
4682 	fs_put_dax(dax_dev);
4683 	return err ? err : ret;
4684 }
4685 
4686 /*
4687  * Setup any per-fs journal parameters now.  We'll do this both on
4688  * initial mount, once the journal has been initialised but before we've
4689  * done any recovery; and again on any subsequent remount.
4690  */
ext4_init_journal_params(struct super_block * sb,journal_t * journal)4691 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4692 {
4693 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4694 
4695 	journal->j_commit_interval = sbi->s_commit_interval;
4696 	journal->j_min_batch_time = sbi->s_min_batch_time;
4697 	journal->j_max_batch_time = sbi->s_max_batch_time;
4698 
4699 	write_lock(&journal->j_state_lock);
4700 	if (test_opt(sb, BARRIER))
4701 		journal->j_flags |= JBD2_BARRIER;
4702 	else
4703 		journal->j_flags &= ~JBD2_BARRIER;
4704 	if (test_opt(sb, DATA_ERR_ABORT))
4705 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4706 	else
4707 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4708 	write_unlock(&journal->j_state_lock);
4709 }
4710 
ext4_get_journal_inode(struct super_block * sb,unsigned int journal_inum)4711 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4712 					     unsigned int journal_inum)
4713 {
4714 	struct inode *journal_inode;
4715 
4716 	/*
4717 	 * Test for the existence of a valid inode on disk.  Bad things
4718 	 * happen if we iget() an unused inode, as the subsequent iput()
4719 	 * will try to delete it.
4720 	 */
4721 	journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
4722 	if (IS_ERR(journal_inode)) {
4723 		ext4_msg(sb, KERN_ERR, "no journal found");
4724 		return NULL;
4725 	}
4726 	if (!journal_inode->i_nlink) {
4727 		make_bad_inode(journal_inode);
4728 		iput(journal_inode);
4729 		ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4730 		return NULL;
4731 	}
4732 
4733 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4734 		  journal_inode, journal_inode->i_size);
4735 	if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) {
4736 		ext4_msg(sb, KERN_ERR, "invalid journal inode");
4737 		iput(journal_inode);
4738 		return NULL;
4739 	}
4740 	return journal_inode;
4741 }
4742 
ext4_get_journal(struct super_block * sb,unsigned int journal_inum)4743 static journal_t *ext4_get_journal(struct super_block *sb,
4744 				   unsigned int journal_inum)
4745 {
4746 	struct inode *journal_inode;
4747 	journal_t *journal;
4748 
4749 	if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
4750 		return NULL;
4751 
4752 	journal_inode = ext4_get_journal_inode(sb, journal_inum);
4753 	if (!journal_inode)
4754 		return NULL;
4755 
4756 	journal = jbd2_journal_init_inode(journal_inode);
4757 	if (!journal) {
4758 		ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4759 		iput(journal_inode);
4760 		return NULL;
4761 	}
4762 	journal->j_private = sb;
4763 	ext4_init_journal_params(sb, journal);
4764 	return journal;
4765 }
4766 
ext4_get_dev_journal(struct super_block * sb,dev_t j_dev)4767 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4768 				       dev_t j_dev)
4769 {
4770 	struct buffer_head *bh;
4771 	journal_t *journal;
4772 	ext4_fsblk_t start;
4773 	ext4_fsblk_t len;
4774 	int hblock, blocksize;
4775 	ext4_fsblk_t sb_block;
4776 	unsigned long offset;
4777 	struct ext4_super_block *es;
4778 	struct block_device *bdev;
4779 
4780 	if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
4781 		return NULL;
4782 
4783 	bdev = ext4_blkdev_get(j_dev, sb);
4784 	if (bdev == NULL)
4785 		return NULL;
4786 
4787 	blocksize = sb->s_blocksize;
4788 	hblock = bdev_logical_block_size(bdev);
4789 	if (blocksize < hblock) {
4790 		ext4_msg(sb, KERN_ERR,
4791 			"blocksize too small for journal device");
4792 		goto out_bdev;
4793 	}
4794 
4795 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4796 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4797 	set_blocksize(bdev, blocksize);
4798 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
4799 		ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4800 		       "external journal");
4801 		goto out_bdev;
4802 	}
4803 
4804 	es = (struct ext4_super_block *) (bh->b_data + offset);
4805 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4806 	    !(le32_to_cpu(es->s_feature_incompat) &
4807 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4808 		ext4_msg(sb, KERN_ERR, "external journal has "
4809 					"bad superblock");
4810 		brelse(bh);
4811 		goto out_bdev;
4812 	}
4813 
4814 	if ((le32_to_cpu(es->s_feature_ro_compat) &
4815 	     EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4816 	    es->s_checksum != ext4_superblock_csum(sb, es)) {
4817 		ext4_msg(sb, KERN_ERR, "external journal has "
4818 				       "corrupt superblock");
4819 		brelse(bh);
4820 		goto out_bdev;
4821 	}
4822 
4823 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4824 		ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4825 		brelse(bh);
4826 		goto out_bdev;
4827 	}
4828 
4829 	len = ext4_blocks_count(es);
4830 	start = sb_block + 1;
4831 	brelse(bh);	/* we're done with the superblock */
4832 
4833 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4834 					start, len, blocksize);
4835 	if (!journal) {
4836 		ext4_msg(sb, KERN_ERR, "failed to create device journal");
4837 		goto out_bdev;
4838 	}
4839 	journal->j_private = sb;
4840 	ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4841 	wait_on_buffer(journal->j_sb_buffer);
4842 	if (!buffer_uptodate(journal->j_sb_buffer)) {
4843 		ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4844 		goto out_journal;
4845 	}
4846 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4847 		ext4_msg(sb, KERN_ERR, "External journal has more than one "
4848 					"user (unsupported) - %d",
4849 			be32_to_cpu(journal->j_superblock->s_nr_users));
4850 		goto out_journal;
4851 	}
4852 	EXT4_SB(sb)->s_journal_bdev = bdev;
4853 	ext4_init_journal_params(sb, journal);
4854 	return journal;
4855 
4856 out_journal:
4857 	jbd2_journal_destroy(journal);
4858 out_bdev:
4859 	ext4_blkdev_put(bdev);
4860 	return NULL;
4861 }
4862 
ext4_load_journal(struct super_block * sb,struct ext4_super_block * es,unsigned long journal_devnum)4863 static int ext4_load_journal(struct super_block *sb,
4864 			     struct ext4_super_block *es,
4865 			     unsigned long journal_devnum)
4866 {
4867 	journal_t *journal;
4868 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4869 	dev_t journal_dev;
4870 	int err = 0;
4871 	int really_read_only;
4872 	int journal_dev_ro;
4873 
4874 	if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
4875 		return -EFSCORRUPTED;
4876 
4877 	if (journal_devnum &&
4878 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4879 		ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4880 			"numbers have changed");
4881 		journal_dev = new_decode_dev(journal_devnum);
4882 	} else
4883 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4884 
4885 	if (journal_inum && journal_dev) {
4886 		ext4_msg(sb, KERN_ERR,
4887 			 "filesystem has both journal inode and journal device!");
4888 		return -EINVAL;
4889 	}
4890 
4891 	if (journal_inum) {
4892 		journal = ext4_get_journal(sb, journal_inum);
4893 		if (!journal)
4894 			return -EINVAL;
4895 	} else {
4896 		journal = ext4_get_dev_journal(sb, journal_dev);
4897 		if (!journal)
4898 			return -EINVAL;
4899 	}
4900 
4901 	journal_dev_ro = bdev_read_only(journal->j_dev);
4902 	really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
4903 
4904 	if (journal_dev_ro && !sb_rdonly(sb)) {
4905 		ext4_msg(sb, KERN_ERR,
4906 			 "journal device read-only, try mounting with '-o ro'");
4907 		err = -EROFS;
4908 		goto err_out;
4909 	}
4910 
4911 	/*
4912 	 * Are we loading a blank journal or performing recovery after a
4913 	 * crash?  For recovery, we need to check in advance whether we
4914 	 * can get read-write access to the device.
4915 	 */
4916 	if (ext4_has_feature_journal_needs_recovery(sb)) {
4917 		if (sb_rdonly(sb)) {
4918 			ext4_msg(sb, KERN_INFO, "INFO: recovery "
4919 					"required on readonly filesystem");
4920 			if (really_read_only) {
4921 				ext4_msg(sb, KERN_ERR, "write access "
4922 					"unavailable, cannot proceed "
4923 					"(try mounting with noload)");
4924 				err = -EROFS;
4925 				goto err_out;
4926 			}
4927 			ext4_msg(sb, KERN_INFO, "write access will "
4928 			       "be enabled during recovery");
4929 		}
4930 	}
4931 
4932 	if (!(journal->j_flags & JBD2_BARRIER))
4933 		ext4_msg(sb, KERN_INFO, "barriers disabled");
4934 
4935 	if (!ext4_has_feature_journal_needs_recovery(sb))
4936 		err = jbd2_journal_wipe(journal, !really_read_only);
4937 	if (!err) {
4938 		char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4939 		if (save)
4940 			memcpy(save, ((char *) es) +
4941 			       EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4942 		err = jbd2_journal_load(journal);
4943 		if (save)
4944 			memcpy(((char *) es) + EXT4_S_ERR_START,
4945 			       save, EXT4_S_ERR_LEN);
4946 		kfree(save);
4947 	}
4948 
4949 	if (err) {
4950 		ext4_msg(sb, KERN_ERR, "error loading journal");
4951 		goto err_out;
4952 	}
4953 
4954 	EXT4_SB(sb)->s_journal = journal;
4955 	err = ext4_clear_journal_err(sb, es);
4956 	if (err) {
4957 		EXT4_SB(sb)->s_journal = NULL;
4958 		jbd2_journal_destroy(journal);
4959 		return err;
4960 	}
4961 
4962 	if (!really_read_only && journal_devnum &&
4963 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4964 		es->s_journal_dev = cpu_to_le32(journal_devnum);
4965 
4966 		/* Make sure we flush the recovery flag to disk. */
4967 		ext4_commit_super(sb, 1);
4968 	}
4969 
4970 	return 0;
4971 
4972 err_out:
4973 	jbd2_journal_destroy(journal);
4974 	return err;
4975 }
4976 
ext4_commit_super(struct super_block * sb,int sync)4977 static int ext4_commit_super(struct super_block *sb, int sync)
4978 {
4979 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4980 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4981 	int error = 0;
4982 
4983 	if (!sbh)
4984 		return -EINVAL;
4985 	if (block_device_ejected(sb))
4986 		return -ENODEV;
4987 
4988 	/*
4989 	 * If the file system is mounted read-only, don't update the
4990 	 * superblock write time.  This avoids updating the superblock
4991 	 * write time when we are mounting the root file system
4992 	 * read/only but we need to replay the journal; at that point,
4993 	 * for people who are east of GMT and who make their clock
4994 	 * tick in localtime for Windows bug-for-bug compatibility,
4995 	 * the clock is set in the future, and this will cause e2fsck
4996 	 * to complain and force a full file system check.
4997 	 */
4998 	if (!(sb->s_flags & SB_RDONLY))
4999 		ext4_update_tstamp(es, s_wtime);
5000 	if (sb->s_bdev->bd_part)
5001 		es->s_kbytes_written =
5002 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
5003 			    ((part_stat_read(sb->s_bdev->bd_part,
5004 					     sectors[STAT_WRITE]) -
5005 			      EXT4_SB(sb)->s_sectors_written_start) >> 1));
5006 	else
5007 		es->s_kbytes_written =
5008 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
5009 	if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5010 		ext4_free_blocks_count_set(es,
5011 			EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5012 				&EXT4_SB(sb)->s_freeclusters_counter)));
5013 	if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5014 		es->s_free_inodes_count =
5015 			cpu_to_le32(percpu_counter_sum_positive(
5016 				&EXT4_SB(sb)->s_freeinodes_counter));
5017 	BUFFER_TRACE(sbh, "marking dirty");
5018 	ext4_superblock_csum_set(sb);
5019 	if (sync)
5020 		lock_buffer(sbh);
5021 	if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5022 		/*
5023 		 * Oh, dear.  A previous attempt to write the
5024 		 * superblock failed.  This could happen because the
5025 		 * USB device was yanked out.  Or it could happen to
5026 		 * be a transient write error and maybe the block will
5027 		 * be remapped.  Nothing we can do but to retry the
5028 		 * write and hope for the best.
5029 		 */
5030 		ext4_msg(sb, KERN_ERR, "previous I/O error to "
5031 		       "superblock detected");
5032 		clear_buffer_write_io_error(sbh);
5033 		set_buffer_uptodate(sbh);
5034 	}
5035 	mark_buffer_dirty(sbh);
5036 	if (sync) {
5037 		unlock_buffer(sbh);
5038 		error = __sync_dirty_buffer(sbh,
5039 			REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5040 		if (buffer_write_io_error(sbh)) {
5041 			ext4_msg(sb, KERN_ERR, "I/O error while writing "
5042 			       "superblock");
5043 			clear_buffer_write_io_error(sbh);
5044 			set_buffer_uptodate(sbh);
5045 		}
5046 	}
5047 	return error;
5048 }
5049 
5050 /*
5051  * Have we just finished recovery?  If so, and if we are mounting (or
5052  * remounting) the filesystem readonly, then we will end up with a
5053  * consistent fs on disk.  Record that fact.
5054  */
ext4_mark_recovery_complete(struct super_block * sb,struct ext4_super_block * es)5055 static int ext4_mark_recovery_complete(struct super_block *sb,
5056 				       struct ext4_super_block *es)
5057 {
5058 	int err;
5059 	journal_t *journal = EXT4_SB(sb)->s_journal;
5060 
5061 	if (!ext4_has_feature_journal(sb)) {
5062 		if (journal != NULL) {
5063 			ext4_error(sb, "Journal got removed while the fs was "
5064 				   "mounted!");
5065 			return -EFSCORRUPTED;
5066 		}
5067 		return 0;
5068 	}
5069 	jbd2_journal_lock_updates(journal);
5070 	err = jbd2_journal_flush(journal);
5071 	if (err < 0)
5072 		goto out;
5073 
5074 	if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5075 		ext4_clear_feature_journal_needs_recovery(sb);
5076 		ext4_commit_super(sb, 1);
5077 	}
5078 out:
5079 	jbd2_journal_unlock_updates(journal);
5080 	return err;
5081 }
5082 
5083 /*
5084  * If we are mounting (or read-write remounting) a filesystem whose journal
5085  * has recorded an error from a previous lifetime, move that error to the
5086  * main filesystem now.
5087  */
ext4_clear_journal_err(struct super_block * sb,struct ext4_super_block * es)5088 static int ext4_clear_journal_err(struct super_block *sb,
5089 				   struct ext4_super_block *es)
5090 {
5091 	journal_t *journal;
5092 	int j_errno;
5093 	const char *errstr;
5094 
5095 	if (!ext4_has_feature_journal(sb)) {
5096 		ext4_error(sb, "Journal got removed while the fs was mounted!");
5097 		return -EFSCORRUPTED;
5098 	}
5099 
5100 	journal = EXT4_SB(sb)->s_journal;
5101 
5102 	/*
5103 	 * Now check for any error status which may have been recorded in the
5104 	 * journal by a prior ext4_error() or ext4_abort()
5105 	 */
5106 
5107 	j_errno = jbd2_journal_errno(journal);
5108 	if (j_errno) {
5109 		char nbuf[16];
5110 
5111 		errstr = ext4_decode_error(sb, j_errno, nbuf);
5112 		ext4_warning(sb, "Filesystem error recorded "
5113 			     "from previous mount: %s", errstr);
5114 		ext4_warning(sb, "Marking fs in need of filesystem check.");
5115 
5116 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5117 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5118 		ext4_commit_super(sb, 1);
5119 
5120 		jbd2_journal_clear_err(journal);
5121 		jbd2_journal_update_sb_errno(journal);
5122 	}
5123 	return 0;
5124 }
5125 
5126 /*
5127  * Force the running and committing transactions to commit,
5128  * and wait on the commit.
5129  */
ext4_force_commit(struct super_block * sb)5130 int ext4_force_commit(struct super_block *sb)
5131 {
5132 	journal_t *journal;
5133 
5134 	if (sb_rdonly(sb))
5135 		return 0;
5136 
5137 	journal = EXT4_SB(sb)->s_journal;
5138 	return ext4_journal_force_commit(journal);
5139 }
5140 
ext4_sync_fs(struct super_block * sb,int wait)5141 static int ext4_sync_fs(struct super_block *sb, int wait)
5142 {
5143 	int ret = 0;
5144 	tid_t target;
5145 	bool needs_barrier = false;
5146 	struct ext4_sb_info *sbi = EXT4_SB(sb);
5147 
5148 	if (unlikely(ext4_forced_shutdown(sbi)))
5149 		return 0;
5150 
5151 	trace_ext4_sync_fs(sb, wait);
5152 	flush_workqueue(sbi->rsv_conversion_wq);
5153 	/*
5154 	 * Writeback quota in non-journalled quota case - journalled quota has
5155 	 * no dirty dquots
5156 	 */
5157 	dquot_writeback_dquots(sb, -1);
5158 	/*
5159 	 * Data writeback is possible w/o journal transaction, so barrier must
5160 	 * being sent at the end of the function. But we can skip it if
5161 	 * transaction_commit will do it for us.
5162 	 */
5163 	if (sbi->s_journal) {
5164 		target = jbd2_get_latest_transaction(sbi->s_journal);
5165 		if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5166 		    !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5167 			needs_barrier = true;
5168 
5169 		if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5170 			if (wait)
5171 				ret = jbd2_log_wait_commit(sbi->s_journal,
5172 							   target);
5173 		}
5174 	} else if (wait && test_opt(sb, BARRIER))
5175 		needs_barrier = true;
5176 	if (needs_barrier) {
5177 		int err;
5178 		err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
5179 		if (!ret)
5180 			ret = err;
5181 	}
5182 
5183 	return ret;
5184 }
5185 
5186 /*
5187  * LVM calls this function before a (read-only) snapshot is created.  This
5188  * gives us a chance to flush the journal completely and mark the fs clean.
5189  *
5190  * Note that only this function cannot bring a filesystem to be in a clean
5191  * state independently. It relies on upper layer to stop all data & metadata
5192  * modifications.
5193  */
ext4_freeze(struct super_block * sb)5194 static int ext4_freeze(struct super_block *sb)
5195 {
5196 	int error = 0;
5197 	journal_t *journal;
5198 
5199 	if (sb_rdonly(sb))
5200 		return 0;
5201 
5202 	journal = EXT4_SB(sb)->s_journal;
5203 
5204 	if (journal) {
5205 		/* Now we set up the journal barrier. */
5206 		jbd2_journal_lock_updates(journal);
5207 
5208 		/*
5209 		 * Don't clear the needs_recovery flag if we failed to
5210 		 * flush the journal.
5211 		 */
5212 		error = jbd2_journal_flush(journal);
5213 		if (error < 0)
5214 			goto out;
5215 
5216 		/* Journal blocked and flushed, clear needs_recovery flag. */
5217 		ext4_clear_feature_journal_needs_recovery(sb);
5218 	}
5219 
5220 	error = ext4_commit_super(sb, 1);
5221 out:
5222 	if (journal)
5223 		/* we rely on upper layer to stop further updates */
5224 		jbd2_journal_unlock_updates(journal);
5225 	return error;
5226 }
5227 
5228 /*
5229  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
5230  * flag here, even though the filesystem is not technically dirty yet.
5231  */
ext4_unfreeze(struct super_block * sb)5232 static int ext4_unfreeze(struct super_block *sb)
5233 {
5234 	if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5235 		return 0;
5236 
5237 	if (EXT4_SB(sb)->s_journal) {
5238 		/* Reset the needs_recovery flag before the fs is unlocked. */
5239 		ext4_set_feature_journal_needs_recovery(sb);
5240 	}
5241 
5242 	ext4_commit_super(sb, 1);
5243 	return 0;
5244 }
5245 
5246 /*
5247  * Structure to save mount options for ext4_remount's benefit
5248  */
5249 struct ext4_mount_options {
5250 	unsigned long s_mount_opt;
5251 	unsigned long s_mount_opt2;
5252 	kuid_t s_resuid;
5253 	kgid_t s_resgid;
5254 	unsigned long s_commit_interval;
5255 	u32 s_min_batch_time, s_max_batch_time;
5256 #ifdef CONFIG_QUOTA
5257 	int s_jquota_fmt;
5258 	char *s_qf_names[EXT4_MAXQUOTAS];
5259 #endif
5260 };
5261 
ext4_remount(struct super_block * sb,int * flags,char * data)5262 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5263 {
5264 	struct ext4_super_block *es;
5265 	struct ext4_sb_info *sbi = EXT4_SB(sb);
5266 	unsigned long old_sb_flags, vfs_flags;
5267 	struct ext4_mount_options old_opts;
5268 	int enable_quota = 0;
5269 	ext4_group_t g;
5270 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5271 	int err = 0;
5272 #ifdef CONFIG_QUOTA
5273 	int i, j;
5274 	char *to_free[EXT4_MAXQUOTAS];
5275 #endif
5276 	char *orig_data = kstrdup(data, GFP_KERNEL);
5277 
5278 	if (data && !orig_data)
5279 		return -ENOMEM;
5280 
5281 	/* Store the original options */
5282 	old_sb_flags = sb->s_flags;
5283 	old_opts.s_mount_opt = sbi->s_mount_opt;
5284 	old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5285 	old_opts.s_resuid = sbi->s_resuid;
5286 	old_opts.s_resgid = sbi->s_resgid;
5287 	old_opts.s_commit_interval = sbi->s_commit_interval;
5288 	old_opts.s_min_batch_time = sbi->s_min_batch_time;
5289 	old_opts.s_max_batch_time = sbi->s_max_batch_time;
5290 #ifdef CONFIG_QUOTA
5291 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5292 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
5293 		if (sbi->s_qf_names[i]) {
5294 			char *qf_name = get_qf_name(sb, sbi, i);
5295 
5296 			old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5297 			if (!old_opts.s_qf_names[i]) {
5298 				for (j = 0; j < i; j++)
5299 					kfree(old_opts.s_qf_names[j]);
5300 				kfree(orig_data);
5301 				return -ENOMEM;
5302 			}
5303 		} else
5304 			old_opts.s_qf_names[i] = NULL;
5305 #endif
5306 	if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5307 		journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5308 
5309 	/*
5310 	 * Some options can be enabled by ext4 and/or by VFS mount flag
5311 	 * either way we need to make sure it matches in both *flags and
5312 	 * s_flags. Copy those selected flags from *flags to s_flags
5313 	 */
5314 	vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5315 	sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5316 
5317 	if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5318 		err = -EINVAL;
5319 		goto restore_opts;
5320 	}
5321 
5322 	if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5323 	    test_opt(sb, JOURNAL_CHECKSUM)) {
5324 		ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5325 			 "during remount not supported; ignoring");
5326 		sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5327 	}
5328 
5329 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5330 		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5331 			ext4_msg(sb, KERN_ERR, "can't mount with "
5332 				 "both data=journal and delalloc");
5333 			err = -EINVAL;
5334 			goto restore_opts;
5335 		}
5336 		if (test_opt(sb, DIOREAD_NOLOCK)) {
5337 			ext4_msg(sb, KERN_ERR, "can't mount with "
5338 				 "both data=journal and dioread_nolock");
5339 			err = -EINVAL;
5340 			goto restore_opts;
5341 		}
5342 	} else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5343 		if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5344 			ext4_msg(sb, KERN_ERR, "can't mount with "
5345 				"journal_async_commit in data=ordered mode");
5346 			err = -EINVAL;
5347 			goto restore_opts;
5348 		}
5349 	}
5350 
5351 	if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5352 		ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5353 		err = -EINVAL;
5354 		goto restore_opts;
5355 	}
5356 
5357 	if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5358 		ext4_abort(sb, "Abort forced by user");
5359 
5360 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5361 		(test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5362 
5363 	es = sbi->s_es;
5364 
5365 	if (sbi->s_journal) {
5366 		ext4_init_journal_params(sb, sbi->s_journal);
5367 		set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5368 	}
5369 
5370 	if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5371 		if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5372 			err = -EROFS;
5373 			goto restore_opts;
5374 		}
5375 
5376 		if (*flags & SB_RDONLY) {
5377 			err = sync_filesystem(sb);
5378 			if (err < 0)
5379 				goto restore_opts;
5380 			err = dquot_suspend(sb, -1);
5381 			if (err < 0)
5382 				goto restore_opts;
5383 
5384 			/*
5385 			 * First of all, the unconditional stuff we have to do
5386 			 * to disable replay of the journal when we next remount
5387 			 */
5388 			sb->s_flags |= SB_RDONLY;
5389 
5390 			/*
5391 			 * OK, test if we are remounting a valid rw partition
5392 			 * readonly, and if so set the rdonly flag and then
5393 			 * mark the partition as valid again.
5394 			 */
5395 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5396 			    (sbi->s_mount_state & EXT4_VALID_FS))
5397 				es->s_state = cpu_to_le16(sbi->s_mount_state);
5398 
5399 			if (sbi->s_journal) {
5400 				/*
5401 				 * We let remount-ro finish even if marking fs
5402 				 * as clean failed...
5403 				 */
5404 				ext4_mark_recovery_complete(sb, es);
5405 			}
5406 			if (sbi->s_mmp_tsk)
5407 				kthread_stop(sbi->s_mmp_tsk);
5408 		} else {
5409 			/* Make sure we can mount this feature set readwrite */
5410 			if (ext4_has_feature_readonly(sb) ||
5411 			    !ext4_feature_set_ok(sb, 0)) {
5412 				err = -EROFS;
5413 				goto restore_opts;
5414 			}
5415 			/*
5416 			 * Make sure the group descriptor checksums
5417 			 * are sane.  If they aren't, refuse to remount r/w.
5418 			 */
5419 			for (g = 0; g < sbi->s_groups_count; g++) {
5420 				struct ext4_group_desc *gdp =
5421 					ext4_get_group_desc(sb, g, NULL);
5422 
5423 				if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5424 					ext4_msg(sb, KERN_ERR,
5425 	       "ext4_remount: Checksum for group %u failed (%u!=%u)",
5426 		g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5427 					       le16_to_cpu(gdp->bg_checksum));
5428 					err = -EFSBADCRC;
5429 					goto restore_opts;
5430 				}
5431 			}
5432 
5433 			/*
5434 			 * If we have an unprocessed orphan list hanging
5435 			 * around from a previously readonly bdev mount,
5436 			 * require a full umount/remount for now.
5437 			 */
5438 			if (es->s_last_orphan) {
5439 				ext4_msg(sb, KERN_WARNING, "Couldn't "
5440 				       "remount RDWR because of unprocessed "
5441 				       "orphan inode list.  Please "
5442 				       "umount/remount instead");
5443 				err = -EINVAL;
5444 				goto restore_opts;
5445 			}
5446 
5447 			/*
5448 			 * Mounting a RDONLY partition read-write, so reread
5449 			 * and store the current valid flag.  (It may have
5450 			 * been changed by e2fsck since we originally mounted
5451 			 * the partition.)
5452 			 */
5453 			if (sbi->s_journal) {
5454 				err = ext4_clear_journal_err(sb, es);
5455 				if (err)
5456 					goto restore_opts;
5457 			}
5458 			sbi->s_mount_state = le16_to_cpu(es->s_state);
5459 
5460 			err = ext4_setup_super(sb, es, 0);
5461 			if (err)
5462 				goto restore_opts;
5463 
5464 			sb->s_flags &= ~SB_RDONLY;
5465 			if (ext4_has_feature_mmp(sb))
5466 				if (ext4_multi_mount_protect(sb,
5467 						le64_to_cpu(es->s_mmp_block))) {
5468 					err = -EROFS;
5469 					goto restore_opts;
5470 				}
5471 			enable_quota = 1;
5472 		}
5473 	}
5474 
5475 	/*
5476 	 * Reinitialize lazy itable initialization thread based on
5477 	 * current settings
5478 	 */
5479 	if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5480 		ext4_unregister_li_request(sb);
5481 	else {
5482 		ext4_group_t first_not_zeroed;
5483 		first_not_zeroed = ext4_has_uninit_itable(sb);
5484 		ext4_register_li_request(sb, first_not_zeroed);
5485 	}
5486 
5487 	/*
5488 	 * Handle creation of system zone data early because it can fail.
5489 	 * Releasing of existing data is done when we are sure remount will
5490 	 * succeed.
5491 	 */
5492 	if (test_opt(sb, BLOCK_VALIDITY) && !sbi->system_blks) {
5493 		err = ext4_setup_system_zone(sb);
5494 		if (err)
5495 			goto restore_opts;
5496 	}
5497 
5498 	if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5499 		err = ext4_commit_super(sb, 1);
5500 		if (err)
5501 			goto restore_opts;
5502 	}
5503 
5504 #ifdef CONFIG_QUOTA
5505 	if (enable_quota) {
5506 		if (sb_any_quota_suspended(sb))
5507 			dquot_resume(sb, -1);
5508 		else if (ext4_has_feature_quota(sb)) {
5509 			err = ext4_enable_quotas(sb);
5510 			if (err)
5511 				goto restore_opts;
5512 		}
5513 	}
5514 	/* Release old quota file names */
5515 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
5516 		kfree(old_opts.s_qf_names[i]);
5517 #endif
5518 	if (!test_opt(sb, BLOCK_VALIDITY) && sbi->system_blks)
5519 		ext4_release_system_zone(sb);
5520 
5521 	/*
5522 	 * Some options can be enabled by ext4 and/or by VFS mount flag
5523 	 * either way we need to make sure it matches in both *flags and
5524 	 * s_flags. Copy those selected flags from s_flags to *flags
5525 	 */
5526 	*flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
5527 
5528 	ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5529 	kfree(orig_data);
5530 	return 0;
5531 
5532 restore_opts:
5533 	/*
5534 	 * If there was a failing r/w to ro transition, we may need to
5535 	 * re-enable quota
5536 	 */
5537 	if ((sb->s_flags & SB_RDONLY) && !(old_sb_flags & SB_RDONLY) &&
5538 	    sb_any_quota_suspended(sb))
5539 		dquot_resume(sb, -1);
5540 	sb->s_flags = old_sb_flags;
5541 	sbi->s_mount_opt = old_opts.s_mount_opt;
5542 	sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5543 	sbi->s_resuid = old_opts.s_resuid;
5544 	sbi->s_resgid = old_opts.s_resgid;
5545 	sbi->s_commit_interval = old_opts.s_commit_interval;
5546 	sbi->s_min_batch_time = old_opts.s_min_batch_time;
5547 	sbi->s_max_batch_time = old_opts.s_max_batch_time;
5548 	if (!test_opt(sb, BLOCK_VALIDITY) && sbi->system_blks)
5549 		ext4_release_system_zone(sb);
5550 #ifdef CONFIG_QUOTA
5551 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5552 	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5553 		to_free[i] = get_qf_name(sb, sbi, i);
5554 		rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
5555 	}
5556 	synchronize_rcu();
5557 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
5558 		kfree(to_free[i]);
5559 #endif
5560 	kfree(orig_data);
5561 	return err;
5562 }
5563 
5564 #ifdef CONFIG_QUOTA
ext4_statfs_project(struct super_block * sb,kprojid_t projid,struct kstatfs * buf)5565 static int ext4_statfs_project(struct super_block *sb,
5566 			       kprojid_t projid, struct kstatfs *buf)
5567 {
5568 	struct kqid qid;
5569 	struct dquot *dquot;
5570 	u64 limit;
5571 	u64 curblock;
5572 
5573 	qid = make_kqid_projid(projid);
5574 	dquot = dqget(sb, qid);
5575 	if (IS_ERR(dquot))
5576 		return PTR_ERR(dquot);
5577 	spin_lock(&dquot->dq_dqb_lock);
5578 
5579 	limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5580 		 dquot->dq_dqb.dqb_bsoftlimit :
5581 		 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5582 	if (limit && buf->f_blocks > limit) {
5583 		curblock = (dquot->dq_dqb.dqb_curspace +
5584 			    dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
5585 		buf->f_blocks = limit;
5586 		buf->f_bfree = buf->f_bavail =
5587 			(buf->f_blocks > curblock) ?
5588 			 (buf->f_blocks - curblock) : 0;
5589 	}
5590 
5591 	limit = dquot->dq_dqb.dqb_isoftlimit ?
5592 		dquot->dq_dqb.dqb_isoftlimit :
5593 		dquot->dq_dqb.dqb_ihardlimit;
5594 	if (limit && buf->f_files > limit) {
5595 		buf->f_files = limit;
5596 		buf->f_ffree =
5597 			(buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5598 			 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5599 	}
5600 
5601 	spin_unlock(&dquot->dq_dqb_lock);
5602 	dqput(dquot);
5603 	return 0;
5604 }
5605 #endif
5606 
ext4_statfs(struct dentry * dentry,struct kstatfs * buf)5607 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5608 {
5609 	struct super_block *sb = dentry->d_sb;
5610 	struct ext4_sb_info *sbi = EXT4_SB(sb);
5611 	struct ext4_super_block *es = sbi->s_es;
5612 	ext4_fsblk_t overhead = 0, resv_blocks;
5613 	u64 fsid;
5614 	s64 bfree;
5615 	resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5616 
5617 	if (!test_opt(sb, MINIX_DF))
5618 		overhead = sbi->s_overhead;
5619 
5620 	buf->f_type = EXT4_SUPER_MAGIC;
5621 	buf->f_bsize = sb->s_blocksize;
5622 	buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5623 	bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5624 		percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5625 	/* prevent underflow in case that few free space is available */
5626 	buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5627 	buf->f_bavail = buf->f_bfree -
5628 			(ext4_r_blocks_count(es) + resv_blocks);
5629 	if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5630 		buf->f_bavail = 0;
5631 	buf->f_files = le32_to_cpu(es->s_inodes_count);
5632 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5633 	buf->f_namelen = EXT4_NAME_LEN;
5634 	fsid = le64_to_cpup((void *)es->s_uuid) ^
5635 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5636 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5637 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5638 
5639 #ifdef CONFIG_QUOTA
5640 	if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5641 	    sb_has_quota_limits_enabled(sb, PRJQUOTA))
5642 		ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5643 #endif
5644 	return 0;
5645 }
5646 
5647 
5648 #ifdef CONFIG_QUOTA
5649 
5650 /*
5651  * Helper functions so that transaction is started before we acquire dqio_sem
5652  * to keep correct lock ordering of transaction > dqio_sem
5653  */
dquot_to_inode(struct dquot * dquot)5654 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5655 {
5656 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5657 }
5658 
ext4_write_dquot(struct dquot * dquot)5659 static int ext4_write_dquot(struct dquot *dquot)
5660 {
5661 	int ret, err;
5662 	handle_t *handle;
5663 	struct inode *inode;
5664 
5665 	inode = dquot_to_inode(dquot);
5666 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5667 				    EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5668 	if (IS_ERR(handle))
5669 		return PTR_ERR(handle);
5670 	ret = dquot_commit(dquot);
5671 	err = ext4_journal_stop(handle);
5672 	if (!ret)
5673 		ret = err;
5674 	return ret;
5675 }
5676 
ext4_acquire_dquot(struct dquot * dquot)5677 static int ext4_acquire_dquot(struct dquot *dquot)
5678 {
5679 	int ret, err;
5680 	handle_t *handle;
5681 
5682 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5683 				    EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5684 	if (IS_ERR(handle))
5685 		return PTR_ERR(handle);
5686 	ret = dquot_acquire(dquot);
5687 	err = ext4_journal_stop(handle);
5688 	if (!ret)
5689 		ret = err;
5690 	return ret;
5691 }
5692 
ext4_release_dquot(struct dquot * dquot)5693 static int ext4_release_dquot(struct dquot *dquot)
5694 {
5695 	int ret, err;
5696 	handle_t *handle;
5697 
5698 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5699 				    EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5700 	if (IS_ERR(handle)) {
5701 		/* Release dquot anyway to avoid endless cycle in dqput() */
5702 		dquot_release(dquot);
5703 		return PTR_ERR(handle);
5704 	}
5705 	ret = dquot_release(dquot);
5706 	err = ext4_journal_stop(handle);
5707 	if (!ret)
5708 		ret = err;
5709 	return ret;
5710 }
5711 
ext4_mark_dquot_dirty(struct dquot * dquot)5712 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5713 {
5714 	struct super_block *sb = dquot->dq_sb;
5715 	struct ext4_sb_info *sbi = EXT4_SB(sb);
5716 
5717 	/* Are we journaling quotas? */
5718 	if (ext4_has_feature_quota(sb) ||
5719 	    sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5720 		dquot_mark_dquot_dirty(dquot);
5721 		return ext4_write_dquot(dquot);
5722 	} else {
5723 		return dquot_mark_dquot_dirty(dquot);
5724 	}
5725 }
5726 
ext4_write_info(struct super_block * sb,int type)5727 static int ext4_write_info(struct super_block *sb, int type)
5728 {
5729 	int ret, err;
5730 	handle_t *handle;
5731 
5732 	/* Data block + inode block */
5733 	handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2);
5734 	if (IS_ERR(handle))
5735 		return PTR_ERR(handle);
5736 	ret = dquot_commit_info(sb, type);
5737 	err = ext4_journal_stop(handle);
5738 	if (!ret)
5739 		ret = err;
5740 	return ret;
5741 }
5742 
5743 /*
5744  * Turn on quotas during mount time - we need to find
5745  * the quota file and such...
5746  */
ext4_quota_on_mount(struct super_block * sb,int type)5747 static int ext4_quota_on_mount(struct super_block *sb, int type)
5748 {
5749 	return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
5750 					EXT4_SB(sb)->s_jquota_fmt, type);
5751 }
5752 
lockdep_set_quota_inode(struct inode * inode,int subclass)5753 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5754 {
5755 	struct ext4_inode_info *ei = EXT4_I(inode);
5756 
5757 	/* The first argument of lockdep_set_subclass has to be
5758 	 * *exactly* the same as the argument to init_rwsem() --- in
5759 	 * this case, in init_once() --- or lockdep gets unhappy
5760 	 * because the name of the lock is set using the
5761 	 * stringification of the argument to init_rwsem().
5762 	 */
5763 	(void) ei;	/* shut up clang warning if !CONFIG_LOCKDEP */
5764 	lockdep_set_subclass(&ei->i_data_sem, subclass);
5765 }
5766 
5767 /*
5768  * Standard function to be called on quota_on
5769  */
ext4_quota_on(struct super_block * sb,int type,int format_id,const struct path * path)5770 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5771 			 const struct path *path)
5772 {
5773 	int err;
5774 
5775 	if (!test_opt(sb, QUOTA))
5776 		return -EINVAL;
5777 
5778 	/* Quotafile not on the same filesystem? */
5779 	if (path->dentry->d_sb != sb)
5780 		return -EXDEV;
5781 
5782 	/* Quota already enabled for this file? */
5783 	if (IS_NOQUOTA(d_inode(path->dentry)))
5784 		return -EBUSY;
5785 
5786 	/* Journaling quota? */
5787 	if (EXT4_SB(sb)->s_qf_names[type]) {
5788 		/* Quotafile not in fs root? */
5789 		if (path->dentry->d_parent != sb->s_root)
5790 			ext4_msg(sb, KERN_WARNING,
5791 				"Quota file not on filesystem root. "
5792 				"Journaled quota will not work");
5793 		sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5794 	} else {
5795 		/*
5796 		 * Clear the flag just in case mount options changed since
5797 		 * last time.
5798 		 */
5799 		sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
5800 	}
5801 
5802 	/*
5803 	 * When we journal data on quota file, we have to flush journal to see
5804 	 * all updates to the file when we bypass pagecache...
5805 	 */
5806 	if (EXT4_SB(sb)->s_journal &&
5807 	    ext4_should_journal_data(d_inode(path->dentry))) {
5808 		/*
5809 		 * We don't need to lock updates but journal_flush() could
5810 		 * otherwise be livelocked...
5811 		 */
5812 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5813 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5814 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5815 		if (err)
5816 			return err;
5817 	}
5818 
5819 	lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5820 	err = dquot_quota_on(sb, type, format_id, path);
5821 	if (!err) {
5822 		struct inode *inode = d_inode(path->dentry);
5823 		handle_t *handle;
5824 
5825 		/*
5826 		 * Set inode flags to prevent userspace from messing with quota
5827 		 * files. If this fails, we return success anyway since quotas
5828 		 * are already enabled and this is not a hard failure.
5829 		 */
5830 		inode_lock(inode);
5831 		handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5832 		if (IS_ERR(handle))
5833 			goto unlock_inode;
5834 		EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5835 		inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5836 				S_NOATIME | S_IMMUTABLE);
5837 		ext4_mark_inode_dirty(handle, inode);
5838 		ext4_journal_stop(handle);
5839 	unlock_inode:
5840 		inode_unlock(inode);
5841 		if (err)
5842 			dquot_quota_off(sb, type);
5843 	}
5844 	if (err)
5845 		lockdep_set_quota_inode(path->dentry->d_inode,
5846 					     I_DATA_SEM_NORMAL);
5847 	return err;
5848 }
5849 
ext4_check_quota_inum(int type,unsigned long qf_inum)5850 static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum)
5851 {
5852 	switch (type) {
5853 	case USRQUOTA:
5854 		return qf_inum == EXT4_USR_QUOTA_INO;
5855 	case GRPQUOTA:
5856 		return qf_inum == EXT4_GRP_QUOTA_INO;
5857 	case PRJQUOTA:
5858 		return qf_inum >= EXT4_GOOD_OLD_FIRST_INO;
5859 	default:
5860 		BUG();
5861 	}
5862 }
5863 
ext4_quota_enable(struct super_block * sb,int type,int format_id,unsigned int flags)5864 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5865 			     unsigned int flags)
5866 {
5867 	int err;
5868 	struct inode *qf_inode;
5869 	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5870 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5871 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5872 		le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5873 	};
5874 
5875 	BUG_ON(!ext4_has_feature_quota(sb));
5876 
5877 	if (!qf_inums[type])
5878 		return -EPERM;
5879 
5880 	if (!ext4_check_quota_inum(type, qf_inums[type])) {
5881 		ext4_error(sb, "Bad quota inum: %lu, type: %d",
5882 				qf_inums[type], type);
5883 		return -EUCLEAN;
5884 	}
5885 
5886 	qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
5887 	if (IS_ERR(qf_inode)) {
5888 		ext4_error(sb, "Bad quota inode: %lu, type: %d",
5889 				qf_inums[type], type);
5890 		return PTR_ERR(qf_inode);
5891 	}
5892 
5893 	/* Don't account quota for quota files to avoid recursion */
5894 	qf_inode->i_flags |= S_NOQUOTA;
5895 	lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5896 	err = dquot_enable(qf_inode, type, format_id, flags);
5897 	if (err)
5898 		lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5899 	iput(qf_inode);
5900 
5901 	return err;
5902 }
5903 
5904 /* Enable usage tracking for all quota types. */
ext4_enable_quotas(struct super_block * sb)5905 static int ext4_enable_quotas(struct super_block *sb)
5906 {
5907 	int type, err = 0;
5908 	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5909 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5910 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5911 		le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5912 	};
5913 	bool quota_mopt[EXT4_MAXQUOTAS] = {
5914 		test_opt(sb, USRQUOTA),
5915 		test_opt(sb, GRPQUOTA),
5916 		test_opt(sb, PRJQUOTA),
5917 	};
5918 
5919 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
5920 	for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5921 		if (qf_inums[type]) {
5922 			err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5923 				DQUOT_USAGE_ENABLED |
5924 				(quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5925 			if (err) {
5926 				ext4_warning(sb,
5927 					"Failed to enable quota tracking "
5928 					"(type=%d, err=%d, ino=%lu). "
5929 					"Please run e2fsck to fix.", type,
5930 					err, qf_inums[type]);
5931 				for (type--; type >= 0; type--) {
5932 					struct inode *inode;
5933 
5934 					inode = sb_dqopt(sb)->files[type];
5935 					if (inode)
5936 						inode = igrab(inode);
5937 					dquot_quota_off(sb, type);
5938 					if (inode) {
5939 						lockdep_set_quota_inode(inode,
5940 							I_DATA_SEM_NORMAL);
5941 						iput(inode);
5942 					}
5943 				}
5944 
5945 				return err;
5946 			}
5947 		}
5948 	}
5949 	return 0;
5950 }
5951 
ext4_quota_off(struct super_block * sb,int type)5952 static int ext4_quota_off(struct super_block *sb, int type)
5953 {
5954 	struct inode *inode = sb_dqopt(sb)->files[type];
5955 	handle_t *handle;
5956 	int err;
5957 
5958 	/* Force all delayed allocation blocks to be allocated.
5959 	 * Caller already holds s_umount sem */
5960 	if (test_opt(sb, DELALLOC))
5961 		sync_filesystem(sb);
5962 
5963 	if (!inode || !igrab(inode))
5964 		goto out;
5965 
5966 	err = dquot_quota_off(sb, type);
5967 	if (err || ext4_has_feature_quota(sb))
5968 		goto out_put;
5969 
5970 	inode_lock(inode);
5971 	/*
5972 	 * Update modification times of quota files when userspace can
5973 	 * start looking at them. If we fail, we return success anyway since
5974 	 * this is not a hard failure and quotas are already disabled.
5975 	 */
5976 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5977 	if (IS_ERR(handle))
5978 		goto out_unlock;
5979 	EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5980 	inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
5981 	inode->i_mtime = inode->i_ctime = current_time(inode);
5982 	ext4_mark_inode_dirty(handle, inode);
5983 	ext4_journal_stop(handle);
5984 out_unlock:
5985 	inode_unlock(inode);
5986 out_put:
5987 	lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
5988 	iput(inode);
5989 	return err;
5990 out:
5991 	return dquot_quota_off(sb, type);
5992 }
5993 
5994 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5995  * acquiring the locks... As quota files are never truncated and quota code
5996  * itself serializes the operations (and no one else should touch the files)
5997  * we don't have to be afraid of races */
ext4_quota_read(struct super_block * sb,int type,char * data,size_t len,loff_t off)5998 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5999 			       size_t len, loff_t off)
6000 {
6001 	struct inode *inode = sb_dqopt(sb)->files[type];
6002 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6003 	int offset = off & (sb->s_blocksize - 1);
6004 	int tocopy;
6005 	size_t toread;
6006 	struct buffer_head *bh;
6007 	loff_t i_size = i_size_read(inode);
6008 
6009 	if (off > i_size)
6010 		return 0;
6011 	if (off+len > i_size)
6012 		len = i_size-off;
6013 	toread = len;
6014 	while (toread > 0) {
6015 		tocopy = sb->s_blocksize - offset < toread ?
6016 				sb->s_blocksize - offset : toread;
6017 		bh = ext4_bread(NULL, inode, blk, 0);
6018 		if (IS_ERR(bh))
6019 			return PTR_ERR(bh);
6020 		if (!bh)	/* A hole? */
6021 			memset(data, 0, tocopy);
6022 		else
6023 			memcpy(data, bh->b_data+offset, tocopy);
6024 		brelse(bh);
6025 		offset = 0;
6026 		toread -= tocopy;
6027 		data += tocopy;
6028 		blk++;
6029 	}
6030 	return len;
6031 }
6032 
6033 /* Write to quotafile (we know the transaction is already started and has
6034  * enough credits) */
ext4_quota_write(struct super_block * sb,int type,const char * data,size_t len,loff_t off)6035 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6036 				const char *data, size_t len, loff_t off)
6037 {
6038 	struct inode *inode = sb_dqopt(sb)->files[type];
6039 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6040 	int err, offset = off & (sb->s_blocksize - 1);
6041 	int retries = 0;
6042 	struct buffer_head *bh;
6043 	handle_t *handle = journal_current_handle();
6044 
6045 	if (!handle) {
6046 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6047 			" cancelled because transaction is not started",
6048 			(unsigned long long)off, (unsigned long long)len);
6049 		return -EIO;
6050 	}
6051 	/*
6052 	 * Since we account only one data block in transaction credits,
6053 	 * then it is impossible to cross a block boundary.
6054 	 */
6055 	if (sb->s_blocksize - offset < len) {
6056 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6057 			" cancelled because not block aligned",
6058 			(unsigned long long)off, (unsigned long long)len);
6059 		return -EIO;
6060 	}
6061 
6062 	do {
6063 		bh = ext4_bread(handle, inode, blk,
6064 				EXT4_GET_BLOCKS_CREATE |
6065 				EXT4_GET_BLOCKS_METADATA_NOFAIL);
6066 	} while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
6067 		 ext4_should_retry_alloc(inode->i_sb, &retries));
6068 	if (IS_ERR(bh))
6069 		return PTR_ERR(bh);
6070 	if (!bh)
6071 		goto out;
6072 	BUFFER_TRACE(bh, "get write access");
6073 	err = ext4_journal_get_write_access(handle, bh);
6074 	if (err) {
6075 		brelse(bh);
6076 		return err;
6077 	}
6078 	lock_buffer(bh);
6079 	memcpy(bh->b_data+offset, data, len);
6080 	flush_dcache_page(bh->b_page);
6081 	unlock_buffer(bh);
6082 	err = ext4_handle_dirty_metadata(handle, NULL, bh);
6083 	brelse(bh);
6084 out:
6085 	if (inode->i_size < off + len) {
6086 		i_size_write(inode, off + len);
6087 		EXT4_I(inode)->i_disksize = inode->i_size;
6088 		ext4_mark_inode_dirty(handle, inode);
6089 	}
6090 	return len;
6091 }
6092 
ext4_get_next_id(struct super_block * sb,struct kqid * qid)6093 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
6094 {
6095 	const struct quota_format_ops	*ops;
6096 
6097 	if (!sb_has_quota_loaded(sb, qid->type))
6098 		return -ESRCH;
6099 	ops = sb_dqopt(sb)->ops[qid->type];
6100 	if (!ops || !ops->get_next_id)
6101 		return -ENOSYS;
6102 	return dquot_get_next_id(sb, qid);
6103 }
6104 #endif
6105 
ext4_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)6106 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6107 		       const char *dev_name, void *data)
6108 {
6109 	return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6110 }
6111 
6112 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
register_as_ext2(void)6113 static inline void register_as_ext2(void)
6114 {
6115 	int err = register_filesystem(&ext2_fs_type);
6116 	if (err)
6117 		printk(KERN_WARNING
6118 		       "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6119 }
6120 
unregister_as_ext2(void)6121 static inline void unregister_as_ext2(void)
6122 {
6123 	unregister_filesystem(&ext2_fs_type);
6124 }
6125 
ext2_feature_set_ok(struct super_block * sb)6126 static inline int ext2_feature_set_ok(struct super_block *sb)
6127 {
6128 	if (ext4_has_unknown_ext2_incompat_features(sb))
6129 		return 0;
6130 	if (sb_rdonly(sb))
6131 		return 1;
6132 	if (ext4_has_unknown_ext2_ro_compat_features(sb))
6133 		return 0;
6134 	return 1;
6135 }
6136 #else
register_as_ext2(void)6137 static inline void register_as_ext2(void) { }
unregister_as_ext2(void)6138 static inline void unregister_as_ext2(void) { }
ext2_feature_set_ok(struct super_block * sb)6139 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6140 #endif
6141 
register_as_ext3(void)6142 static inline void register_as_ext3(void)
6143 {
6144 	int err = register_filesystem(&ext3_fs_type);
6145 	if (err)
6146 		printk(KERN_WARNING
6147 		       "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6148 }
6149 
unregister_as_ext3(void)6150 static inline void unregister_as_ext3(void)
6151 {
6152 	unregister_filesystem(&ext3_fs_type);
6153 }
6154 
ext3_feature_set_ok(struct super_block * sb)6155 static inline int ext3_feature_set_ok(struct super_block *sb)
6156 {
6157 	if (ext4_has_unknown_ext3_incompat_features(sb))
6158 		return 0;
6159 	if (!ext4_has_feature_journal(sb))
6160 		return 0;
6161 	if (sb_rdonly(sb))
6162 		return 1;
6163 	if (ext4_has_unknown_ext3_ro_compat_features(sb))
6164 		return 0;
6165 	return 1;
6166 }
6167 
6168 static struct file_system_type ext4_fs_type = {
6169 	.owner		= THIS_MODULE,
6170 	.name		= "ext4",
6171 	.mount		= ext4_mount,
6172 	.kill_sb	= kill_block_super,
6173 	.fs_flags	= FS_REQUIRES_DEV,
6174 };
6175 MODULE_ALIAS_FS("ext4");
6176 
6177 /* Shared across all ext4 file systems */
6178 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6179 
ext4_init_fs(void)6180 static int __init ext4_init_fs(void)
6181 {
6182 	int i, err;
6183 
6184 	ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6185 	ext4_li_info = NULL;
6186 	mutex_init(&ext4_li_mtx);
6187 
6188 	/* Build-time check for flags consistency */
6189 	ext4_check_flag_values();
6190 
6191 	for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6192 		init_waitqueue_head(&ext4__ioend_wq[i]);
6193 
6194 	err = ext4_init_es();
6195 	if (err)
6196 		return err;
6197 
6198 	err = ext4_init_pageio();
6199 	if (err)
6200 		goto out5;
6201 
6202 	err = ext4_init_system_zone();
6203 	if (err)
6204 		goto out4;
6205 
6206 	err = ext4_init_sysfs();
6207 	if (err)
6208 		goto out3;
6209 
6210 	err = ext4_init_mballoc();
6211 	if (err)
6212 		goto out2;
6213 	err = init_inodecache();
6214 	if (err)
6215 		goto out1;
6216 	register_as_ext3();
6217 	register_as_ext2();
6218 	err = register_filesystem(&ext4_fs_type);
6219 	if (err)
6220 		goto out;
6221 
6222 	return 0;
6223 out:
6224 	unregister_as_ext2();
6225 	unregister_as_ext3();
6226 	destroy_inodecache();
6227 out1:
6228 	ext4_exit_mballoc();
6229 out2:
6230 	ext4_exit_sysfs();
6231 out3:
6232 	ext4_exit_system_zone();
6233 out4:
6234 	ext4_exit_pageio();
6235 out5:
6236 	ext4_exit_es();
6237 
6238 	return err;
6239 }
6240 
ext4_exit_fs(void)6241 static void __exit ext4_exit_fs(void)
6242 {
6243 	ext4_destroy_lazyinit_thread();
6244 	unregister_as_ext2();
6245 	unregister_as_ext3();
6246 	unregister_filesystem(&ext4_fs_type);
6247 	destroy_inodecache();
6248 	ext4_exit_mballoc();
6249 	ext4_exit_sysfs();
6250 	ext4_exit_system_zone();
6251 	ext4_exit_pageio();
6252 	ext4_exit_es();
6253 }
6254 
6255 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6256 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6257 MODULE_LICENSE("GPL");
6258 MODULE_SOFTDEP("pre: crc32c");
6259 module_init(ext4_init_fs)
6260 module_exit(ext4_exit_fs)
6261