]>
Commit | Line | Data |
---|---|---|
ac27a0ec | 1 | /* |
617ba13b | 2 | * linux/fs/ext4/super.c |
ac27a0ec DK |
3 | * |
4 | * Copyright (C) 1992, 1993, 1994, 1995 | |
5 | * Remy Card (card@masi.ibp.fr) | |
6 | * Laboratoire MASI - Institut Blaise Pascal | |
7 | * Universite Pierre et Marie Curie (Paris VI) | |
8 | * | |
9 | * from | |
10 | * | |
11 | * linux/fs/minix/inode.c | |
12 | * | |
13 | * Copyright (C) 1991, 1992 Linus Torvalds | |
14 | * | |
15 | * Big-endian to little-endian byte-swapping/bitmaps by | |
16 | * David S. Miller (davem@caip.rutgers.edu), 1995 | |
17 | */ | |
18 | ||
19 | #include <linux/module.h> | |
20 | #include <linux/string.h> | |
21 | #include <linux/fs.h> | |
22 | #include <linux/time.h> | |
c5ca7c76 | 23 | #include <linux/vmalloc.h> |
ac27a0ec DK |
24 | #include <linux/slab.h> |
25 | #include <linux/init.h> | |
26 | #include <linux/blkdev.h> | |
66114cad | 27 | #include <linux/backing-dev.h> |
ac27a0ec | 28 | #include <linux/parser.h> |
ac27a0ec | 29 | #include <linux/buffer_head.h> |
a5694255 | 30 | #include <linux/exportfs.h> |
ac27a0ec DK |
31 | #include <linux/vfs.h> |
32 | #include <linux/random.h> | |
33 | #include <linux/mount.h> | |
34 | #include <linux/namei.h> | |
35 | #include <linux/quotaops.h> | |
36 | #include <linux/seq_file.h> | |
3197ebdb | 37 | #include <linux/ctype.h> |
1330593e | 38 | #include <linux/log2.h> |
717d50e4 | 39 | #include <linux/crc16.h> |
ef510424 | 40 | #include <linux/dax.h> |
7abc52c2 | 41 | #include <linux/cleancache.h> |
7c0f6ba6 | 42 | #include <linux/uaccess.h> |
ac27a0ec | 43 | |
bfff6873 LC |
44 | #include <linux/kthread.h> |
45 | #include <linux/freezer.h> | |
46 | ||
3dcf5451 | 47 | #include "ext4.h" |
4a092d73 | 48 | #include "ext4_extents.h" /* Needed for trace points definition */ |
3dcf5451 | 49 | #include "ext4_jbd2.h" |
ac27a0ec DK |
50 | #include "xattr.h" |
51 | #include "acl.h" | |
3661d286 | 52 | #include "mballoc.h" |
0c9ec4be | 53 | #include "fsmap.h" |
ac27a0ec | 54 | |
9bffad1e TT |
55 | #define CREATE_TRACE_POINTS |
56 | #include <trace/events/ext4.h> | |
57 | ||
0b75a840 LC |
58 | static struct ext4_lazy_init *ext4_li_info; |
59 | static struct mutex ext4_li_mtx; | |
e294a537 | 60 | static struct ratelimit_state ext4_mount_msg_ratelimit; |
9f6200bb | 61 | |
617ba13b | 62 | static int ext4_load_journal(struct super_block *, struct ext4_super_block *, |
ac27a0ec | 63 | unsigned long journal_devnum); |
2adf6da8 | 64 | static int ext4_show_options(struct seq_file *seq, struct dentry *root); |
e2d67052 | 65 | static int ext4_commit_super(struct super_block *sb, int sync); |
2b2d6d01 TT |
66 | static void ext4_mark_recovery_complete(struct super_block *sb, |
67 | struct ext4_super_block *es); | |
68 | static void ext4_clear_journal_err(struct super_block *sb, | |
69 | struct ext4_super_block *es); | |
617ba13b | 70 | static int ext4_sync_fs(struct super_block *sb, int wait); |
2b2d6d01 TT |
71 | static int ext4_remount(struct super_block *sb, int *flags, char *data); |
72 | static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); | |
c4be0c1d | 73 | static int ext4_unfreeze(struct super_block *sb); |
c4be0c1d | 74 | static int ext4_freeze(struct super_block *sb); |
152a0836 AV |
75 | static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, |
76 | const char *dev_name, void *data); | |
2035e776 TT |
77 | static inline int ext2_feature_set_ok(struct super_block *sb); |
78 | static inline int ext3_feature_set_ok(struct super_block *sb); | |
d39195c3 | 79 | static int ext4_feature_set_ok(struct super_block *sb, int readonly); |
bfff6873 LC |
80 | static void ext4_destroy_lazyinit_thread(void); |
81 | static void ext4_unregister_li_request(struct super_block *sb); | |
8f1f7453 | 82 | static void ext4_clear_request_list(void); |
c6cb7e77 EW |
83 | static struct inode *ext4_get_journal_inode(struct super_block *sb, |
84 | unsigned int journal_inum); | |
ac27a0ec | 85 | |
e74031fd JK |
86 | /* |
87 | * Lock ordering | |
88 | * | |
89 | * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and | |
90 | * i_mmap_rwsem (inode->i_mmap_rwsem)! | |
91 | * | |
92 | * page fault path: | |
93 | * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start -> | |
94 | * page lock -> i_data_sem (rw) | |
95 | * | |
96 | * buffered write path: | |
97 | * sb_start_write -> i_mutex -> mmap_sem | |
98 | * sb_start_write -> i_mutex -> transaction start -> page lock -> | |
99 | * i_data_sem (rw) | |
100 | * | |
101 | * truncate: | |
102 | * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) -> | |
103 | * i_mmap_rwsem (w) -> page lock | |
104 | * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) -> | |
105 | * transaction start -> i_data_sem (rw) | |
106 | * | |
107 | * direct IO: | |
108 | * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem | |
109 | * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> | |
110 | * transaction start -> i_data_sem (rw) | |
111 | * | |
112 | * writepages: | |
113 | * transaction start -> page lock(s) -> i_data_sem (rw) | |
114 | */ | |
115 | ||
c290ea01 | 116 | #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) |
2035e776 TT |
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 | }; | |
7f78e035 | 124 | MODULE_ALIAS_FS("ext2"); |
fa7614dd | 125 | MODULE_ALIAS("ext2"); |
2035e776 TT |
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 | ||
ba69f9ab JK |
132 | static struct file_system_type ext3_fs_type = { |
133 | .owner = THIS_MODULE, | |
134 | .name = "ext3", | |
152a0836 | 135 | .mount = ext4_mount, |
ba69f9ab JK |
136 | .kill_sb = kill_block_super, |
137 | .fs_flags = FS_REQUIRES_DEV, | |
138 | }; | |
7f78e035 | 139 | MODULE_ALIAS_FS("ext3"); |
fa7614dd | 140 | MODULE_ALIAS("ext3"); |
ba69f9ab | 141 | #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) |
bd81d8ee | 142 | |
d25425f8 DW |
143 | static int ext4_verify_csum_type(struct super_block *sb, |
144 | struct ext4_super_block *es) | |
145 | { | |
e2b911c5 | 146 | if (!ext4_has_feature_metadata_csum(sb)) |
d25425f8 DW |
147 | return 1; |
148 | ||
149 | return es->s_checksum_type == EXT4_CRC32C_CHKSUM; | |
150 | } | |
151 | ||
a9c47317 DW |
152 | static __le32 ext4_superblock_csum(struct super_block *sb, |
153 | struct ext4_super_block *es) | |
154 | { | |
155 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
156 | int offset = offsetof(struct ext4_super_block, s_checksum); | |
157 | __u32 csum; | |
158 | ||
159 | csum = ext4_chksum(sbi, ~0, (char *)es, offset); | |
160 | ||
161 | return cpu_to_le32(csum); | |
162 | } | |
163 | ||
c197855e SH |
164 | static int ext4_superblock_csum_verify(struct super_block *sb, |
165 | struct ext4_super_block *es) | |
a9c47317 | 166 | { |
9aa5d32b | 167 | if (!ext4_has_metadata_csum(sb)) |
a9c47317 DW |
168 | return 1; |
169 | ||
170 | return es->s_checksum == ext4_superblock_csum(sb, es); | |
171 | } | |
172 | ||
06db49e6 | 173 | void ext4_superblock_csum_set(struct super_block *sb) |
a9c47317 | 174 | { |
06db49e6 TT |
175 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
176 | ||
9aa5d32b | 177 | if (!ext4_has_metadata_csum(sb)) |
a9c47317 DW |
178 | return; |
179 | ||
180 | es->s_checksum = ext4_superblock_csum(sb, es); | |
181 | } | |
182 | ||
9933fc0a TT |
183 | void *ext4_kvmalloc(size_t size, gfp_t flags) |
184 | { | |
185 | void *ret; | |
186 | ||
8be04b93 | 187 | ret = kmalloc(size, flags | __GFP_NOWARN); |
9933fc0a TT |
188 | if (!ret) |
189 | ret = __vmalloc(size, flags, PAGE_KERNEL); | |
190 | return ret; | |
191 | } | |
192 | ||
193 | void *ext4_kvzalloc(size_t size, gfp_t flags) | |
194 | { | |
195 | void *ret; | |
196 | ||
8be04b93 | 197 | ret = kzalloc(size, flags | __GFP_NOWARN); |
9933fc0a TT |
198 | if (!ret) |
199 | ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL); | |
200 | return ret; | |
201 | } | |
202 | ||
8fadc143 AR |
203 | ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, |
204 | struct ext4_group_desc *bg) | |
bd81d8ee | 205 | { |
3a14589c | 206 | return le32_to_cpu(bg->bg_block_bitmap_lo) | |
8fadc143 | 207 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 208 | (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); |
bd81d8ee LV |
209 | } |
210 | ||
8fadc143 AR |
211 | ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, |
212 | struct ext4_group_desc *bg) | |
bd81d8ee | 213 | { |
5272f837 | 214 | return le32_to_cpu(bg->bg_inode_bitmap_lo) | |
8fadc143 | 215 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 216 | (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); |
bd81d8ee LV |
217 | } |
218 | ||
8fadc143 AR |
219 | ext4_fsblk_t ext4_inode_table(struct super_block *sb, |
220 | struct ext4_group_desc *bg) | |
bd81d8ee | 221 | { |
5272f837 | 222 | return le32_to_cpu(bg->bg_inode_table_lo) | |
8fadc143 | 223 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 224 | (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); |
bd81d8ee LV |
225 | } |
226 | ||
021b65bb TT |
227 | __u32 ext4_free_group_clusters(struct super_block *sb, |
228 | struct ext4_group_desc *bg) | |
560671a0 AK |
229 | { |
230 | return le16_to_cpu(bg->bg_free_blocks_count_lo) | | |
231 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 232 | (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); |
560671a0 AK |
233 | } |
234 | ||
235 | __u32 ext4_free_inodes_count(struct super_block *sb, | |
236 | struct ext4_group_desc *bg) | |
237 | { | |
238 | return le16_to_cpu(bg->bg_free_inodes_count_lo) | | |
239 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 240 | (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); |
560671a0 AK |
241 | } |
242 | ||
243 | __u32 ext4_used_dirs_count(struct super_block *sb, | |
244 | struct ext4_group_desc *bg) | |
245 | { | |
246 | return le16_to_cpu(bg->bg_used_dirs_count_lo) | | |
247 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 248 | (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); |
560671a0 AK |
249 | } |
250 | ||
251 | __u32 ext4_itable_unused_count(struct super_block *sb, | |
252 | struct ext4_group_desc *bg) | |
253 | { | |
254 | return le16_to_cpu(bg->bg_itable_unused_lo) | | |
255 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 256 | (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); |
560671a0 AK |
257 | } |
258 | ||
8fadc143 AR |
259 | void ext4_block_bitmap_set(struct super_block *sb, |
260 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 261 | { |
3a14589c | 262 | bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
263 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
264 | bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
265 | } |
266 | ||
8fadc143 AR |
267 | void ext4_inode_bitmap_set(struct super_block *sb, |
268 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 269 | { |
5272f837 | 270 | bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
271 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
272 | bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
273 | } |
274 | ||
8fadc143 AR |
275 | void ext4_inode_table_set(struct super_block *sb, |
276 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 277 | { |
5272f837 | 278 | bg->bg_inode_table_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
279 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
280 | bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
281 | } |
282 | ||
021b65bb TT |
283 | void ext4_free_group_clusters_set(struct super_block *sb, |
284 | struct ext4_group_desc *bg, __u32 count) | |
560671a0 AK |
285 | { |
286 | bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); | |
287 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
288 | bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); | |
289 | } | |
290 | ||
291 | void ext4_free_inodes_set(struct super_block *sb, | |
292 | struct ext4_group_desc *bg, __u32 count) | |
293 | { | |
294 | bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); | |
295 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
296 | bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); | |
297 | } | |
298 | ||
299 | void ext4_used_dirs_set(struct super_block *sb, | |
300 | struct ext4_group_desc *bg, __u32 count) | |
301 | { | |
302 | bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); | |
303 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
304 | bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); | |
305 | } | |
306 | ||
307 | void ext4_itable_unused_set(struct super_block *sb, | |
308 | struct ext4_group_desc *bg, __u32 count) | |
309 | { | |
310 | bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); | |
311 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
312 | bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); | |
313 | } | |
314 | ||
d3d1faf6 | 315 | |
1c13d5c0 TT |
316 | static void __save_error_info(struct super_block *sb, const char *func, |
317 | unsigned int line) | |
318 | { | |
319 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; | |
320 | ||
321 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; | |
1b46617b TT |
322 | if (bdev_read_only(sb->s_bdev)) |
323 | return; | |
1c13d5c0 TT |
324 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); |
325 | es->s_last_error_time = cpu_to_le32(get_seconds()); | |
326 | strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func)); | |
327 | es->s_last_error_line = cpu_to_le32(line); | |
328 | if (!es->s_first_error_time) { | |
329 | es->s_first_error_time = es->s_last_error_time; | |
330 | strncpy(es->s_first_error_func, func, | |
331 | sizeof(es->s_first_error_func)); | |
332 | es->s_first_error_line = cpu_to_le32(line); | |
333 | es->s_first_error_ino = es->s_last_error_ino; | |
334 | es->s_first_error_block = es->s_last_error_block; | |
335 | } | |
66e61a9e TT |
336 | /* |
337 | * Start the daily error reporting function if it hasn't been | |
338 | * started already | |
339 | */ | |
340 | if (!es->s_error_count) | |
341 | mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ); | |
ba39ebb6 | 342 | le32_add_cpu(&es->s_error_count, 1); |
1c13d5c0 TT |
343 | } |
344 | ||
345 | static void save_error_info(struct super_block *sb, const char *func, | |
346 | unsigned int line) | |
347 | { | |
348 | __save_error_info(sb, func, line); | |
349 | ext4_commit_super(sb, 1); | |
350 | } | |
351 | ||
bdfe0cbd TT |
352 | /* |
353 | * The del_gendisk() function uninitializes the disk-specific data | |
354 | * structures, including the bdi structure, without telling anyone | |
355 | * else. Once this happens, any attempt to call mark_buffer_dirty() | |
356 | * (for example, by ext4_commit_super), will cause a kernel OOPS. | |
357 | * This is a kludge to prevent these oops until we can put in a proper | |
358 | * hook in del_gendisk() to inform the VFS and file system layers. | |
359 | */ | |
360 | static int block_device_ejected(struct super_block *sb) | |
361 | { | |
362 | struct inode *bd_inode = sb->s_bdev->bd_inode; | |
363 | struct backing_dev_info *bdi = inode_to_bdi(bd_inode); | |
364 | ||
365 | return bdi->dev == NULL; | |
366 | } | |
367 | ||
18aadd47 BJ |
368 | static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) |
369 | { | |
370 | struct super_block *sb = journal->j_private; | |
371 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
372 | int error = is_journal_aborted(journal); | |
5d3ee208 | 373 | struct ext4_journal_cb_entry *jce; |
18aadd47 | 374 | |
5d3ee208 | 375 | BUG_ON(txn->t_state == T_FINISHED); |
a0154344 DJ |
376 | |
377 | ext4_process_freed_data(sb, txn->t_tid); | |
378 | ||
18aadd47 | 379 | spin_lock(&sbi->s_md_lock); |
5d3ee208 DM |
380 | while (!list_empty(&txn->t_private_list)) { |
381 | jce = list_entry(txn->t_private_list.next, | |
382 | struct ext4_journal_cb_entry, jce_list); | |
18aadd47 BJ |
383 | list_del_init(&jce->jce_list); |
384 | spin_unlock(&sbi->s_md_lock); | |
385 | jce->jce_func(sb, jce, error); | |
386 | spin_lock(&sbi->s_md_lock); | |
387 | } | |
388 | spin_unlock(&sbi->s_md_lock); | |
389 | } | |
1c13d5c0 | 390 | |
ac27a0ec DK |
391 | /* Deal with the reporting of failure conditions on a filesystem such as |
392 | * inconsistencies detected or read IO failures. | |
393 | * | |
394 | * On ext2, we can store the error state of the filesystem in the | |
617ba13b | 395 | * superblock. That is not possible on ext4, because we may have other |
ac27a0ec DK |
396 | * write ordering constraints on the superblock which prevent us from |
397 | * writing it out straight away; and given that the journal is about to | |
398 | * be aborted, we can't rely on the current, or future, transactions to | |
399 | * write out the superblock safely. | |
400 | * | |
dab291af | 401 | * We'll just use the jbd2_journal_abort() error code to record an error in |
d6b198bc | 402 | * the journal instead. On recovery, the journal will complain about |
ac27a0ec DK |
403 | * that error until we've noted it down and cleared it. |
404 | */ | |
405 | ||
617ba13b | 406 | static void ext4_handle_error(struct super_block *sb) |
ac27a0ec | 407 | { |
bc98a42c | 408 | if (sb_rdonly(sb)) |
ac27a0ec DK |
409 | return; |
410 | ||
2b2d6d01 | 411 | if (!test_opt(sb, ERRORS_CONT)) { |
617ba13b | 412 | journal_t *journal = EXT4_SB(sb)->s_journal; |
ac27a0ec | 413 | |
4ab2f15b | 414 | EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; |
ac27a0ec | 415 | if (journal) |
dab291af | 416 | jbd2_journal_abort(journal, -EIO); |
ac27a0ec | 417 | } |
2b2d6d01 | 418 | if (test_opt(sb, ERRORS_RO)) { |
b31e1552 | 419 | ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); |
4418e141 DM |
420 | /* |
421 | * Make sure updated value of ->s_mount_flags will be visible | |
422 | * before ->s_flags update | |
423 | */ | |
424 | smp_wmb(); | |
ac27a0ec DK |
425 | sb->s_flags |= MS_RDONLY; |
426 | } | |
4327ba52 DJ |
427 | if (test_opt(sb, ERRORS_PANIC)) { |
428 | if (EXT4_SB(sb)->s_journal && | |
429 | !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR)) | |
430 | return; | |
617ba13b | 431 | panic("EXT4-fs (device %s): panic forced after error\n", |
ac27a0ec | 432 | sb->s_id); |
4327ba52 | 433 | } |
ac27a0ec DK |
434 | } |
435 | ||
efbed4dc TT |
436 | #define ext4_error_ratelimit(sb) \ |
437 | ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \ | |
438 | "EXT4-fs error") | |
439 | ||
12062ddd | 440 | void __ext4_error(struct super_block *sb, const char *function, |
c398eda0 | 441 | unsigned int line, const char *fmt, ...) |
ac27a0ec | 442 | { |
0ff2ea7d | 443 | struct va_format vaf; |
ac27a0ec DK |
444 | va_list args; |
445 | ||
0db1ff22 TT |
446 | if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) |
447 | return; | |
448 | ||
efbed4dc TT |
449 | if (ext4_error_ratelimit(sb)) { |
450 | va_start(args, fmt); | |
451 | vaf.fmt = fmt; | |
452 | vaf.va = &args; | |
453 | printk(KERN_CRIT | |
454 | "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", | |
455 | sb->s_id, function, line, current->comm, &vaf); | |
456 | va_end(args); | |
457 | } | |
f3fc0210 | 458 | save_error_info(sb, function, line); |
617ba13b | 459 | ext4_handle_error(sb); |
ac27a0ec DK |
460 | } |
461 | ||
e7c96e8e JP |
462 | void __ext4_error_inode(struct inode *inode, const char *function, |
463 | unsigned int line, ext4_fsblk_t block, | |
464 | const char *fmt, ...) | |
273df556 FM |
465 | { |
466 | va_list args; | |
f7c21177 | 467 | struct va_format vaf; |
1c13d5c0 | 468 | struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es; |
273df556 | 469 | |
0db1ff22 TT |
470 | if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) |
471 | return; | |
472 | ||
1c13d5c0 TT |
473 | es->s_last_error_ino = cpu_to_le32(inode->i_ino); |
474 | es->s_last_error_block = cpu_to_le64(block); | |
efbed4dc TT |
475 | if (ext4_error_ratelimit(inode->i_sb)) { |
476 | va_start(args, fmt); | |
477 | vaf.fmt = fmt; | |
478 | vaf.va = &args; | |
479 | if (block) | |
480 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " | |
481 | "inode #%lu: block %llu: comm %s: %pV\n", | |
482 | inode->i_sb->s_id, function, line, inode->i_ino, | |
483 | block, current->comm, &vaf); | |
484 | else | |
485 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " | |
486 | "inode #%lu: comm %s: %pV\n", | |
487 | inode->i_sb->s_id, function, line, inode->i_ino, | |
488 | current->comm, &vaf); | |
489 | va_end(args); | |
490 | } | |
1c13d5c0 | 491 | save_error_info(inode->i_sb, function, line); |
273df556 FM |
492 | ext4_handle_error(inode->i_sb); |
493 | } | |
494 | ||
e7c96e8e JP |
495 | void __ext4_error_file(struct file *file, const char *function, |
496 | unsigned int line, ext4_fsblk_t block, | |
497 | const char *fmt, ...) | |
273df556 FM |
498 | { |
499 | va_list args; | |
f7c21177 | 500 | struct va_format vaf; |
1c13d5c0 | 501 | struct ext4_super_block *es; |
496ad9aa | 502 | struct inode *inode = file_inode(file); |
273df556 FM |
503 | char pathname[80], *path; |
504 | ||
0db1ff22 TT |
505 | if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) |
506 | return; | |
507 | ||
1c13d5c0 TT |
508 | es = EXT4_SB(inode->i_sb)->s_es; |
509 | es->s_last_error_ino = cpu_to_le32(inode->i_ino); | |
efbed4dc | 510 | if (ext4_error_ratelimit(inode->i_sb)) { |
9bf39ab2 | 511 | path = file_path(file, pathname, sizeof(pathname)); |
efbed4dc TT |
512 | if (IS_ERR(path)) |
513 | path = "(unknown)"; | |
514 | va_start(args, fmt); | |
515 | vaf.fmt = fmt; | |
516 | vaf.va = &args; | |
517 | if (block) | |
518 | printk(KERN_CRIT | |
519 | "EXT4-fs error (device %s): %s:%d: inode #%lu: " | |
520 | "block %llu: comm %s: path %s: %pV\n", | |
521 | inode->i_sb->s_id, function, line, inode->i_ino, | |
522 | block, current->comm, path, &vaf); | |
523 | else | |
524 | printk(KERN_CRIT | |
525 | "EXT4-fs error (device %s): %s:%d: inode #%lu: " | |
526 | "comm %s: path %s: %pV\n", | |
527 | inode->i_sb->s_id, function, line, inode->i_ino, | |
528 | current->comm, path, &vaf); | |
529 | va_end(args); | |
530 | } | |
1c13d5c0 | 531 | save_error_info(inode->i_sb, function, line); |
273df556 FM |
532 | ext4_handle_error(inode->i_sb); |
533 | } | |
534 | ||
722887dd TT |
535 | const char *ext4_decode_error(struct super_block *sb, int errno, |
536 | char nbuf[16]) | |
ac27a0ec DK |
537 | { |
538 | char *errstr = NULL; | |
539 | ||
540 | switch (errno) { | |
6a797d27 DW |
541 | case -EFSCORRUPTED: |
542 | errstr = "Corrupt filesystem"; | |
543 | break; | |
544 | case -EFSBADCRC: | |
545 | errstr = "Filesystem failed CRC"; | |
546 | break; | |
ac27a0ec DK |
547 | case -EIO: |
548 | errstr = "IO failure"; | |
549 | break; | |
550 | case -ENOMEM: | |
551 | errstr = "Out of memory"; | |
552 | break; | |
553 | case -EROFS: | |
78f1ddbb TT |
554 | if (!sb || (EXT4_SB(sb)->s_journal && |
555 | EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) | |
ac27a0ec DK |
556 | errstr = "Journal has aborted"; |
557 | else | |
558 | errstr = "Readonly filesystem"; | |
559 | break; | |
560 | default: | |
561 | /* If the caller passed in an extra buffer for unknown | |
562 | * errors, textualise them now. Else we just return | |
563 | * NULL. */ | |
564 | if (nbuf) { | |
565 | /* Check for truncated error codes... */ | |
566 | if (snprintf(nbuf, 16, "error %d", -errno) >= 0) | |
567 | errstr = nbuf; | |
568 | } | |
569 | break; | |
570 | } | |
571 | ||
572 | return errstr; | |
573 | } | |
574 | ||
617ba13b | 575 | /* __ext4_std_error decodes expected errors from journaling functions |
ac27a0ec DK |
576 | * automatically and invokes the appropriate error response. */ |
577 | ||
c398eda0 TT |
578 | void __ext4_std_error(struct super_block *sb, const char *function, |
579 | unsigned int line, int errno) | |
ac27a0ec DK |
580 | { |
581 | char nbuf[16]; | |
582 | const char *errstr; | |
583 | ||
0db1ff22 TT |
584 | if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) |
585 | return; | |
586 | ||
ac27a0ec DK |
587 | /* Special case: if the error is EROFS, and we're not already |
588 | * inside a transaction, then there's really no point in logging | |
589 | * an error. */ | |
bc98a42c | 590 | if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb)) |
ac27a0ec DK |
591 | return; |
592 | ||
efbed4dc TT |
593 | if (ext4_error_ratelimit(sb)) { |
594 | errstr = ext4_decode_error(sb, errno, nbuf); | |
595 | printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", | |
596 | sb->s_id, function, line, errstr); | |
597 | } | |
ac27a0ec | 598 | |
efbed4dc | 599 | save_error_info(sb, function, line); |
617ba13b | 600 | ext4_handle_error(sb); |
ac27a0ec DK |
601 | } |
602 | ||
603 | /* | |
617ba13b | 604 | * ext4_abort is a much stronger failure handler than ext4_error. The |
ac27a0ec DK |
605 | * abort function may be used to deal with unrecoverable failures such |
606 | * as journal IO errors or ENOMEM at a critical moment in log management. | |
607 | * | |
608 | * We unconditionally force the filesystem into an ABORT|READONLY state, | |
609 | * unless the error response on the fs has been set to panic in which | |
610 | * case we take the easy way out and panic immediately. | |
611 | */ | |
612 | ||
c67d859e | 613 | void __ext4_abort(struct super_block *sb, const char *function, |
c398eda0 | 614 | unsigned int line, const char *fmt, ...) |
ac27a0ec | 615 | { |
651e1c3b | 616 | struct va_format vaf; |
ac27a0ec DK |
617 | va_list args; |
618 | ||
0db1ff22 TT |
619 | if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) |
620 | return; | |
621 | ||
1c13d5c0 | 622 | save_error_info(sb, function, line); |
ac27a0ec | 623 | va_start(args, fmt); |
651e1c3b JP |
624 | vaf.fmt = fmt; |
625 | vaf.va = &args; | |
626 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n", | |
627 | sb->s_id, function, line, &vaf); | |
ac27a0ec DK |
628 | va_end(args); |
629 | ||
bc98a42c | 630 | if (sb_rdonly(sb) == 0) { |
1c13d5c0 | 631 | ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); |
1c13d5c0 | 632 | EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; |
4418e141 DM |
633 | /* |
634 | * Make sure updated value of ->s_mount_flags will be visible | |
635 | * before ->s_flags update | |
636 | */ | |
637 | smp_wmb(); | |
638 | sb->s_flags |= MS_RDONLY; | |
1c13d5c0 TT |
639 | if (EXT4_SB(sb)->s_journal) |
640 | jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); | |
641 | save_error_info(sb, function, line); | |
642 | } | |
4327ba52 DJ |
643 | if (test_opt(sb, ERRORS_PANIC)) { |
644 | if (EXT4_SB(sb)->s_journal && | |
645 | !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR)) | |
646 | return; | |
617ba13b | 647 | panic("EXT4-fs panic from previous error\n"); |
4327ba52 | 648 | } |
ac27a0ec DK |
649 | } |
650 | ||
e7c96e8e JP |
651 | void __ext4_msg(struct super_block *sb, |
652 | const char *prefix, const char *fmt, ...) | |
b31e1552 | 653 | { |
0ff2ea7d | 654 | struct va_format vaf; |
b31e1552 ES |
655 | va_list args; |
656 | ||
efbed4dc TT |
657 | if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs")) |
658 | return; | |
659 | ||
b31e1552 | 660 | va_start(args, fmt); |
0ff2ea7d JP |
661 | vaf.fmt = fmt; |
662 | vaf.va = &args; | |
663 | printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); | |
b31e1552 ES |
664 | va_end(args); |
665 | } | |
666 | ||
b03a2f7e AD |
667 | #define ext4_warning_ratelimit(sb) \ |
668 | ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \ | |
669 | "EXT4-fs warning") | |
670 | ||
12062ddd | 671 | void __ext4_warning(struct super_block *sb, const char *function, |
c398eda0 | 672 | unsigned int line, const char *fmt, ...) |
ac27a0ec | 673 | { |
0ff2ea7d | 674 | struct va_format vaf; |
ac27a0ec DK |
675 | va_list args; |
676 | ||
b03a2f7e | 677 | if (!ext4_warning_ratelimit(sb)) |
efbed4dc TT |
678 | return; |
679 | ||
ac27a0ec | 680 | va_start(args, fmt); |
0ff2ea7d JP |
681 | vaf.fmt = fmt; |
682 | vaf.va = &args; | |
683 | printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", | |
684 | sb->s_id, function, line, &vaf); | |
ac27a0ec DK |
685 | va_end(args); |
686 | } | |
687 | ||
b03a2f7e AD |
688 | void __ext4_warning_inode(const struct inode *inode, const char *function, |
689 | unsigned int line, const char *fmt, ...) | |
690 | { | |
691 | struct va_format vaf; | |
692 | va_list args; | |
693 | ||
694 | if (!ext4_warning_ratelimit(inode->i_sb)) | |
695 | return; | |
696 | ||
697 | va_start(args, fmt); | |
698 | vaf.fmt = fmt; | |
699 | vaf.va = &args; | |
700 | printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: " | |
701 | "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id, | |
702 | function, line, inode->i_ino, current->comm, &vaf); | |
703 | va_end(args); | |
704 | } | |
705 | ||
e29136f8 TT |
706 | void __ext4_grp_locked_error(const char *function, unsigned int line, |
707 | struct super_block *sb, ext4_group_t grp, | |
708 | unsigned long ino, ext4_fsblk_t block, | |
709 | const char *fmt, ...) | |
5d1b1b3f AK |
710 | __releases(bitlock) |
711 | __acquires(bitlock) | |
712 | { | |
0ff2ea7d | 713 | struct va_format vaf; |
5d1b1b3f AK |
714 | va_list args; |
715 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; | |
716 | ||
0db1ff22 TT |
717 | if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) |
718 | return; | |
719 | ||
1c13d5c0 TT |
720 | es->s_last_error_ino = cpu_to_le32(ino); |
721 | es->s_last_error_block = cpu_to_le64(block); | |
722 | __save_error_info(sb, function, line); | |
0ff2ea7d | 723 | |
efbed4dc TT |
724 | if (ext4_error_ratelimit(sb)) { |
725 | va_start(args, fmt); | |
726 | vaf.fmt = fmt; | |
727 | vaf.va = &args; | |
728 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", | |
729 | sb->s_id, function, line, grp); | |
730 | if (ino) | |
731 | printk(KERN_CONT "inode %lu: ", ino); | |
732 | if (block) | |
733 | printk(KERN_CONT "block %llu:", | |
734 | (unsigned long long) block); | |
735 | printk(KERN_CONT "%pV\n", &vaf); | |
736 | va_end(args); | |
737 | } | |
5d1b1b3f AK |
738 | |
739 | if (test_opt(sb, ERRORS_CONT)) { | |
e2d67052 | 740 | ext4_commit_super(sb, 0); |
5d1b1b3f AK |
741 | return; |
742 | } | |
1c13d5c0 | 743 | |
5d1b1b3f AK |
744 | ext4_unlock_group(sb, grp); |
745 | ext4_handle_error(sb); | |
746 | /* | |
747 | * We only get here in the ERRORS_RO case; relocking the group | |
748 | * may be dangerous, but nothing bad will happen since the | |
749 | * filesystem will have already been marked read/only and the | |
750 | * journal has been aborted. We return 1 as a hint to callers | |
751 | * who might what to use the return value from | |
25985edc | 752 | * ext4_grp_locked_error() to distinguish between the |
5d1b1b3f AK |
753 | * ERRORS_CONT and ERRORS_RO case, and perhaps return more |
754 | * aggressively from the ext4 function in question, with a | |
755 | * more appropriate error code. | |
756 | */ | |
757 | ext4_lock_group(sb, grp); | |
758 | return; | |
759 | } | |
760 | ||
617ba13b | 761 | void ext4_update_dynamic_rev(struct super_block *sb) |
ac27a0ec | 762 | { |
617ba13b | 763 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
ac27a0ec | 764 | |
617ba13b | 765 | if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) |
ac27a0ec DK |
766 | return; |
767 | ||
12062ddd | 768 | ext4_warning(sb, |
ac27a0ec DK |
769 | "updating to rev %d because of new feature flag, " |
770 | "running e2fsck is recommended", | |
617ba13b | 771 | EXT4_DYNAMIC_REV); |
ac27a0ec | 772 | |
617ba13b MC |
773 | es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); |
774 | es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); | |
775 | es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); | |
ac27a0ec DK |
776 | /* leave es->s_feature_*compat flags alone */ |
777 | /* es->s_uuid will be set by e2fsck if empty */ | |
778 | ||
779 | /* | |
780 | * The rest of the superblock fields should be zero, and if not it | |
781 | * means they are likely already in use, so leave them alone. We | |
782 | * can leave it up to e2fsck to clean up any inconsistencies there. | |
783 | */ | |
784 | } | |
785 | ||
786 | /* | |
787 | * Open the external journal device | |
788 | */ | |
b31e1552 | 789 | static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) |
ac27a0ec DK |
790 | { |
791 | struct block_device *bdev; | |
792 | char b[BDEVNAME_SIZE]; | |
793 | ||
d4d77629 | 794 | bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); |
ac27a0ec DK |
795 | if (IS_ERR(bdev)) |
796 | goto fail; | |
797 | return bdev; | |
798 | ||
799 | fail: | |
b31e1552 | 800 | ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld", |
ac27a0ec DK |
801 | __bdevname(dev, b), PTR_ERR(bdev)); |
802 | return NULL; | |
803 | } | |
804 | ||
805 | /* | |
806 | * Release the journal device | |
807 | */ | |
4385bab1 | 808 | static void ext4_blkdev_put(struct block_device *bdev) |
ac27a0ec | 809 | { |
4385bab1 | 810 | blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); |
ac27a0ec DK |
811 | } |
812 | ||
4385bab1 | 813 | static void ext4_blkdev_remove(struct ext4_sb_info *sbi) |
ac27a0ec DK |
814 | { |
815 | struct block_device *bdev; | |
ac27a0ec DK |
816 | bdev = sbi->journal_bdev; |
817 | if (bdev) { | |
4385bab1 | 818 | ext4_blkdev_put(bdev); |
ac27a0ec DK |
819 | sbi->journal_bdev = NULL; |
820 | } | |
ac27a0ec DK |
821 | } |
822 | ||
823 | static inline struct inode *orphan_list_entry(struct list_head *l) | |
824 | { | |
617ba13b | 825 | return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; |
ac27a0ec DK |
826 | } |
827 | ||
617ba13b | 828 | static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) |
ac27a0ec DK |
829 | { |
830 | struct list_head *l; | |
831 | ||
b31e1552 ES |
832 | ext4_msg(sb, KERN_ERR, "sb orphan head is %d", |
833 | le32_to_cpu(sbi->s_es->s_last_orphan)); | |
ac27a0ec DK |
834 | |
835 | printk(KERN_ERR "sb_info orphan list:\n"); | |
836 | list_for_each(l, &sbi->s_orphan) { | |
837 | struct inode *inode = orphan_list_entry(l); | |
838 | printk(KERN_ERR " " | |
839 | "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", | |
840 | inode->i_sb->s_id, inode->i_ino, inode, | |
841 | inode->i_mode, inode->i_nlink, | |
842 | NEXT_ORPHAN(inode)); | |
843 | } | |
844 | } | |
845 | ||
957153fc JK |
846 | #ifdef CONFIG_QUOTA |
847 | static int ext4_quota_off(struct super_block *sb, int type); | |
848 | ||
849 | static inline void ext4_quota_off_umount(struct super_block *sb) | |
850 | { | |
851 | int type; | |
852 | ||
964edf66 JK |
853 | /* Use our quota_off function to clear inode flags etc. */ |
854 | for (type = 0; type < EXT4_MAXQUOTAS; type++) | |
855 | ext4_quota_off(sb, type); | |
957153fc JK |
856 | } |
857 | #else | |
858 | static inline void ext4_quota_off_umount(struct super_block *sb) | |
859 | { | |
860 | } | |
861 | #endif | |
862 | ||
2b2d6d01 | 863 | static void ext4_put_super(struct super_block *sb) |
ac27a0ec | 864 | { |
617ba13b MC |
865 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
866 | struct ext4_super_block *es = sbi->s_es; | |
97abd7d4 | 867 | int aborted = 0; |
ef2cabf7 | 868 | int i, err; |
ac27a0ec | 869 | |
857ac889 | 870 | ext4_unregister_li_request(sb); |
957153fc | 871 | ext4_quota_off_umount(sb); |
e0ccfd95 | 872 | |
2e8fa54e | 873 | flush_workqueue(sbi->rsv_conversion_wq); |
2e8fa54e | 874 | destroy_workqueue(sbi->rsv_conversion_wq); |
4c0425ff | 875 | |
0390131b | 876 | if (sbi->s_journal) { |
97abd7d4 | 877 | aborted = is_journal_aborted(sbi->s_journal); |
0390131b FM |
878 | err = jbd2_journal_destroy(sbi->s_journal); |
879 | sbi->s_journal = NULL; | |
97abd7d4 | 880 | if ((err < 0) && !aborted) |
c67d859e | 881 | ext4_abort(sb, "Couldn't clean up the journal"); |
0390131b | 882 | } |
d4edac31 | 883 | |
ebd173be | 884 | ext4_unregister_sysfs(sb); |
d3922a77 | 885 | ext4_es_unregister_shrinker(sbi); |
9105bb14 | 886 | del_timer_sync(&sbi->s_err_report); |
d4edac31 JB |
887 | ext4_release_system_zone(sb); |
888 | ext4_mb_release(sb); | |
889 | ext4_ext_release(sb); | |
d4edac31 | 890 | |
bc98a42c | 891 | if (!sb_rdonly(sb) && !aborted) { |
e2b911c5 | 892 | ext4_clear_feature_journal_needs_recovery(sb); |
ac27a0ec | 893 | es->s_state = cpu_to_le16(sbi->s_mount_state); |
ac27a0ec | 894 | } |
bc98a42c | 895 | if (!sb_rdonly(sb)) |
a8e25a83 AB |
896 | ext4_commit_super(sb, 1); |
897 | ||
ac27a0ec DK |
898 | for (i = 0; i < sbi->s_gdb_count; i++) |
899 | brelse(sbi->s_group_desc[i]); | |
b93b41d4 AV |
900 | kvfree(sbi->s_group_desc); |
901 | kvfree(sbi->s_flex_groups); | |
57042651 | 902 | percpu_counter_destroy(&sbi->s_freeclusters_counter); |
ac27a0ec DK |
903 | percpu_counter_destroy(&sbi->s_freeinodes_counter); |
904 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
57042651 | 905 | percpu_counter_destroy(&sbi->s_dirtyclusters_counter); |
c8585c6f | 906 | percpu_free_rwsem(&sbi->s_journal_flag_rwsem); |
ac27a0ec | 907 | #ifdef CONFIG_QUOTA |
a2d4a646 | 908 | for (i = 0; i < EXT4_MAXQUOTAS; i++) |
ac27a0ec DK |
909 | kfree(sbi->s_qf_names[i]); |
910 | #endif | |
911 | ||
912 | /* Debugging code just in case the in-memory inode orphan list | |
913 | * isn't empty. The on-disk one can be non-empty if we've | |
914 | * detected an error and taken the fs readonly, but the | |
915 | * in-memory list had better be clean by this point. */ | |
916 | if (!list_empty(&sbi->s_orphan)) | |
917 | dump_orphan_list(sb, sbi); | |
918 | J_ASSERT(list_empty(&sbi->s_orphan)); | |
919 | ||
89d96a6f | 920 | sync_blockdev(sb->s_bdev); |
f98393a6 | 921 | invalidate_bdev(sb->s_bdev); |
ac27a0ec DK |
922 | if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { |
923 | /* | |
924 | * Invalidate the journal device's buffers. We don't want them | |
925 | * floating about in memory - the physical journal device may | |
926 | * hotswapped, and it breaks the `ro-after' testing code. | |
927 | */ | |
928 | sync_blockdev(sbi->journal_bdev); | |
f98393a6 | 929 | invalidate_bdev(sbi->journal_bdev); |
617ba13b | 930 | ext4_blkdev_remove(sbi); |
ac27a0ec | 931 | } |
dec214d0 TE |
932 | if (sbi->s_ea_inode_cache) { |
933 | ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); | |
934 | sbi->s_ea_inode_cache = NULL; | |
935 | } | |
47387409 TE |
936 | if (sbi->s_ea_block_cache) { |
937 | ext4_xattr_destroy_cache(sbi->s_ea_block_cache); | |
938 | sbi->s_ea_block_cache = NULL; | |
9c191f70 | 939 | } |
c5e06d10 JL |
940 | if (sbi->s_mmp_tsk) |
941 | kthread_stop(sbi->s_mmp_tsk); | |
9060dd2c | 942 | brelse(sbi->s_sbh); |
ac27a0ec | 943 | sb->s_fs_info = NULL; |
3197ebdb TT |
944 | /* |
945 | * Now that we are completely done shutting down the | |
946 | * superblock, we need to actually destroy the kobject. | |
947 | */ | |
3197ebdb TT |
948 | kobject_put(&sbi->s_kobj); |
949 | wait_for_completion(&sbi->s_kobj_unregister); | |
0441984a DW |
950 | if (sbi->s_chksum_driver) |
951 | crypto_free_shash(sbi->s_chksum_driver); | |
705895b6 | 952 | kfree(sbi->s_blockgroup_lock); |
ac27a0ec | 953 | kfree(sbi); |
ac27a0ec DK |
954 | } |
955 | ||
e18b890b | 956 | static struct kmem_cache *ext4_inode_cachep; |
ac27a0ec DK |
957 | |
958 | /* | |
959 | * Called inside transaction, so use GFP_NOFS | |
960 | */ | |
617ba13b | 961 | static struct inode *ext4_alloc_inode(struct super_block *sb) |
ac27a0ec | 962 | { |
617ba13b | 963 | struct ext4_inode_info *ei; |
ac27a0ec | 964 | |
e6b4f8da | 965 | ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); |
ac27a0ec DK |
966 | if (!ei) |
967 | return NULL; | |
0b8e58a1 | 968 | |
ac27a0ec | 969 | ei->vfs_inode.i_version = 1; |
202ee5df | 970 | spin_lock_init(&ei->i_raw_lock); |
c9de560d AT |
971 | INIT_LIST_HEAD(&ei->i_prealloc_list); |
972 | spin_lock_init(&ei->i_prealloc_lock); | |
9a26b661 ZL |
973 | ext4_es_init_tree(&ei->i_es_tree); |
974 | rwlock_init(&ei->i_es_lock); | |
edaa53ca | 975 | INIT_LIST_HEAD(&ei->i_es_list); |
eb68d0e2 | 976 | ei->i_es_all_nr = 0; |
edaa53ca | 977 | ei->i_es_shk_nr = 0; |
dd475925 | 978 | ei->i_es_shrink_lblk = 0; |
d2a17637 MC |
979 | ei->i_reserved_data_blocks = 0; |
980 | ei->i_reserved_meta_blocks = 0; | |
981 | ei->i_allocated_meta_blocks = 0; | |
9d0be502 | 982 | ei->i_da_metadata_calc_len = 0; |
7e731bc9 | 983 | ei->i_da_metadata_calc_last_lblock = 0; |
d2a17637 | 984 | spin_lock_init(&(ei->i_block_reservation_lock)); |
a9e7f447 DM |
985 | #ifdef CONFIG_QUOTA |
986 | ei->i_reserved_quota = 0; | |
96c7e0d9 | 987 | memset(&ei->i_dquot, 0, sizeof(ei->i_dquot)); |
a9e7f447 | 988 | #endif |
8aefcd55 | 989 | ei->jinode = NULL; |
2e8fa54e | 990 | INIT_LIST_HEAD(&ei->i_rsv_conversion_list); |
744692dc | 991 | spin_lock_init(&ei->i_completed_io_lock); |
b436b9be JK |
992 | ei->i_sync_tid = 0; |
993 | ei->i_datasync_tid = 0; | |
e27f41e1 | 994 | atomic_set(&ei->i_unwritten, 0); |
2e8fa54e | 995 | INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); |
ac27a0ec DK |
996 | return &ei->vfs_inode; |
997 | } | |
998 | ||
7ff9c073 TT |
999 | static int ext4_drop_inode(struct inode *inode) |
1000 | { | |
1001 | int drop = generic_drop_inode(inode); | |
1002 | ||
1003 | trace_ext4_drop_inode(inode, drop); | |
1004 | return drop; | |
1005 | } | |
1006 | ||
fa0d7e3d NP |
1007 | static void ext4_i_callback(struct rcu_head *head) |
1008 | { | |
1009 | struct inode *inode = container_of(head, struct inode, i_rcu); | |
fa0d7e3d NP |
1010 | kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); |
1011 | } | |
1012 | ||
617ba13b | 1013 | static void ext4_destroy_inode(struct inode *inode) |
ac27a0ec | 1014 | { |
9f7dd93d | 1015 | if (!list_empty(&(EXT4_I(inode)->i_orphan))) { |
b31e1552 ES |
1016 | ext4_msg(inode->i_sb, KERN_ERR, |
1017 | "Inode %lu (%p): orphan list check failed!", | |
1018 | inode->i_ino, EXT4_I(inode)); | |
9f7dd93d VA |
1019 | print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, |
1020 | EXT4_I(inode), sizeof(struct ext4_inode_info), | |
1021 | true); | |
1022 | dump_stack(); | |
1023 | } | |
fa0d7e3d | 1024 | call_rcu(&inode->i_rcu, ext4_i_callback); |
ac27a0ec DK |
1025 | } |
1026 | ||
51cc5068 | 1027 | static void init_once(void *foo) |
ac27a0ec | 1028 | { |
617ba13b | 1029 | struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; |
ac27a0ec | 1030 | |
a35afb83 | 1031 | INIT_LIST_HEAD(&ei->i_orphan); |
a35afb83 | 1032 | init_rwsem(&ei->xattr_sem); |
0e855ac8 | 1033 | init_rwsem(&ei->i_data_sem); |
ea3d7209 | 1034 | init_rwsem(&ei->i_mmap_sem); |
a35afb83 | 1035 | inode_init_once(&ei->vfs_inode); |
ac27a0ec DK |
1036 | } |
1037 | ||
e67bc2b3 | 1038 | static int __init init_inodecache(void) |
ac27a0ec | 1039 | { |
617ba13b MC |
1040 | ext4_inode_cachep = kmem_cache_create("ext4_inode_cache", |
1041 | sizeof(struct ext4_inode_info), | |
ac27a0ec | 1042 | 0, (SLAB_RECLAIM_ACCOUNT| |
5d097056 | 1043 | SLAB_MEM_SPREAD|SLAB_ACCOUNT), |
20c2df83 | 1044 | init_once); |
617ba13b | 1045 | if (ext4_inode_cachep == NULL) |
ac27a0ec DK |
1046 | return -ENOMEM; |
1047 | return 0; | |
1048 | } | |
1049 | ||
1050 | static void destroy_inodecache(void) | |
1051 | { | |
8c0a8537 KS |
1052 | /* |
1053 | * Make sure all delayed rcu free inodes are flushed before we | |
1054 | * destroy cache. | |
1055 | */ | |
1056 | rcu_barrier(); | |
617ba13b | 1057 | kmem_cache_destroy(ext4_inode_cachep); |
ac27a0ec DK |
1058 | } |
1059 | ||
0930fcc1 | 1060 | void ext4_clear_inode(struct inode *inode) |
ac27a0ec | 1061 | { |
0930fcc1 | 1062 | invalidate_inode_buffers(inode); |
dbd5768f | 1063 | clear_inode(inode); |
9f754758 | 1064 | dquot_drop(inode); |
c2ea3fde | 1065 | ext4_discard_preallocations(inode); |
51865fda | 1066 | ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); |
8aefcd55 TT |
1067 | if (EXT4_I(inode)->jinode) { |
1068 | jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), | |
1069 | EXT4_I(inode)->jinode); | |
1070 | jbd2_free_inode(EXT4_I(inode)->jinode); | |
1071 | EXT4_I(inode)->jinode = NULL; | |
1072 | } | |
b7236e21 | 1073 | #ifdef CONFIG_EXT4_FS_ENCRYPTION |
a7550b30 | 1074 | fscrypt_put_encryption_info(inode, NULL); |
b7236e21 | 1075 | #endif |
ac27a0ec DK |
1076 | } |
1077 | ||
1b961ac0 | 1078 | static struct inode *ext4_nfs_get_inode(struct super_block *sb, |
0b8e58a1 | 1079 | u64 ino, u32 generation) |
ac27a0ec | 1080 | { |
ac27a0ec | 1081 | struct inode *inode; |
ac27a0ec | 1082 | |
617ba13b | 1083 | if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO) |
ac27a0ec | 1084 | return ERR_PTR(-ESTALE); |
617ba13b | 1085 | if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count)) |
ac27a0ec DK |
1086 | return ERR_PTR(-ESTALE); |
1087 | ||
1088 | /* iget isn't really right if the inode is currently unallocated!! | |
1089 | * | |
617ba13b | 1090 | * ext4_read_inode will return a bad_inode if the inode had been |
ac27a0ec DK |
1091 | * deleted, so we should be safe. |
1092 | * | |
1093 | * Currently we don't know the generation for parent directory, so | |
1094 | * a generation of 0 means "accept any" | |
1095 | */ | |
f4bb2981 | 1096 | inode = ext4_iget_normal(sb, ino); |
1d1fe1ee DH |
1097 | if (IS_ERR(inode)) |
1098 | return ERR_CAST(inode); | |
1099 | if (generation && inode->i_generation != generation) { | |
ac27a0ec DK |
1100 | iput(inode); |
1101 | return ERR_PTR(-ESTALE); | |
1102 | } | |
1b961ac0 CH |
1103 | |
1104 | return inode; | |
1105 | } | |
1106 | ||
1107 | static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, | |
0b8e58a1 | 1108 | int fh_len, int fh_type) |
1b961ac0 CH |
1109 | { |
1110 | return generic_fh_to_dentry(sb, fid, fh_len, fh_type, | |
1111 | ext4_nfs_get_inode); | |
1112 | } | |
1113 | ||
1114 | static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, | |
0b8e58a1 | 1115 | int fh_len, int fh_type) |
1b961ac0 CH |
1116 | { |
1117 | return generic_fh_to_parent(sb, fid, fh_len, fh_type, | |
1118 | ext4_nfs_get_inode); | |
ac27a0ec DK |
1119 | } |
1120 | ||
c39a7f84 TO |
1121 | /* |
1122 | * Try to release metadata pages (indirect blocks, directories) which are | |
1123 | * mapped via the block device. Since these pages could have journal heads | |
1124 | * which would prevent try_to_free_buffers() from freeing them, we must use | |
1125 | * jbd2 layer's try_to_free_buffers() function to release them. | |
1126 | */ | |
0b8e58a1 AD |
1127 | static int bdev_try_to_free_page(struct super_block *sb, struct page *page, |
1128 | gfp_t wait) | |
c39a7f84 TO |
1129 | { |
1130 | journal_t *journal = EXT4_SB(sb)->s_journal; | |
1131 | ||
1132 | WARN_ON(PageChecked(page)); | |
1133 | if (!page_has_buffers(page)) | |
1134 | return 0; | |
1135 | if (journal) | |
1136 | return jbd2_journal_try_to_free_buffers(journal, page, | |
d0164adc | 1137 | wait & ~__GFP_DIRECT_RECLAIM); |
c39a7f84 TO |
1138 | return try_to_free_buffers(page); |
1139 | } | |
1140 | ||
a7550b30 JK |
1141 | #ifdef CONFIG_EXT4_FS_ENCRYPTION |
1142 | static int ext4_get_context(struct inode *inode, void *ctx, size_t len) | |
1143 | { | |
1144 | return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION, | |
1145 | EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len); | |
1146 | } | |
1147 | ||
a7550b30 JK |
1148 | static int ext4_set_context(struct inode *inode, const void *ctx, size_t len, |
1149 | void *fs_data) | |
1150 | { | |
2f8f5e76 | 1151 | handle_t *handle = fs_data; |
c1a5d5f6 | 1152 | int res, res2, credits, retries = 0; |
2f8f5e76 | 1153 | |
9ce0151a EB |
1154 | /* |
1155 | * Encrypting the root directory is not allowed because e2fsck expects | |
1156 | * lost+found to exist and be unencrypted, and encrypting the root | |
1157 | * directory would imply encrypting the lost+found directory as well as | |
1158 | * the filename "lost+found" itself. | |
1159 | */ | |
1160 | if (inode->i_ino == EXT4_ROOT_INO) | |
1161 | return -EPERM; | |
2f8f5e76 | 1162 | |
94840e3c EB |
1163 | res = ext4_convert_inline_data(inode); |
1164 | if (res) | |
1165 | return res; | |
1166 | ||
2f8f5e76 EB |
1167 | /* |
1168 | * If a journal handle was specified, then the encryption context is | |
1169 | * being set on a new inode via inheritance and is part of a larger | |
1170 | * transaction to create the inode. Otherwise the encryption context is | |
1171 | * being set on an existing inode in its own transaction. Only in the | |
1172 | * latter case should the "retry on ENOSPC" logic be used. | |
1173 | */ | |
a7550b30 | 1174 | |
2f8f5e76 EB |
1175 | if (handle) { |
1176 | res = ext4_xattr_set_handle(handle, inode, | |
1177 | EXT4_XATTR_INDEX_ENCRYPTION, | |
1178 | EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, | |
1179 | ctx, len, 0); | |
a7550b30 JK |
1180 | if (!res) { |
1181 | ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); | |
1182 | ext4_clear_inode_state(inode, | |
1183 | EXT4_STATE_MAY_INLINE_DATA); | |
a3caa24b JK |
1184 | /* |
1185 | * Update inode->i_flags - e.g. S_DAX may get disabled | |
1186 | */ | |
1187 | ext4_set_inode_flags(inode); | |
a7550b30 JK |
1188 | } |
1189 | return res; | |
1190 | } | |
1191 | ||
b8cb5a54 TE |
1192 | res = dquot_initialize(inode); |
1193 | if (res) | |
1194 | return res; | |
2f8f5e76 | 1195 | retry: |
af65207c TE |
1196 | res = ext4_xattr_set_credits(inode, len, false /* is_create */, |
1197 | &credits); | |
dec214d0 TE |
1198 | if (res) |
1199 | return res; | |
1200 | ||
c1a5d5f6 | 1201 | handle = ext4_journal_start(inode, EXT4_HT_MISC, credits); |
a7550b30 JK |
1202 | if (IS_ERR(handle)) |
1203 | return PTR_ERR(handle); | |
1204 | ||
2f8f5e76 EB |
1205 | res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION, |
1206 | EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, | |
1207 | ctx, len, 0); | |
a7550b30 JK |
1208 | if (!res) { |
1209 | ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); | |
a3caa24b JK |
1210 | /* Update inode->i_flags - e.g. S_DAX may get disabled */ |
1211 | ext4_set_inode_flags(inode); | |
a7550b30 JK |
1212 | res = ext4_mark_inode_dirty(handle, inode); |
1213 | if (res) | |
1214 | EXT4_ERROR_INODE(inode, "Failed to mark inode dirty"); | |
1215 | } | |
1216 | res2 = ext4_journal_stop(handle); | |
2f8f5e76 EB |
1217 | |
1218 | if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) | |
1219 | goto retry; | |
a7550b30 JK |
1220 | if (!res) |
1221 | res = res2; | |
1222 | return res; | |
1223 | } | |
1224 | ||
c250b7dd | 1225 | static bool ext4_dummy_context(struct inode *inode) |
a7550b30 JK |
1226 | { |
1227 | return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb)); | |
1228 | } | |
1229 | ||
1230 | static unsigned ext4_max_namelen(struct inode *inode) | |
1231 | { | |
1232 | return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize : | |
1233 | EXT4_NAME_LEN; | |
1234 | } | |
1235 | ||
6f69f0ed | 1236 | static const struct fscrypt_operations ext4_cryptops = { |
a5d431ef | 1237 | .key_prefix = "ext4:", |
a7550b30 | 1238 | .get_context = ext4_get_context, |
a7550b30 JK |
1239 | .set_context = ext4_set_context, |
1240 | .dummy_context = ext4_dummy_context, | |
1241 | .is_encrypted = ext4_encrypted_inode, | |
1242 | .empty_dir = ext4_empty_dir, | |
1243 | .max_namelen = ext4_max_namelen, | |
1244 | }; | |
1245 | #else | |
6f69f0ed | 1246 | static const struct fscrypt_operations ext4_cryptops = { |
a7550b30 JK |
1247 | .is_encrypted = ext4_encrypted_inode, |
1248 | }; | |
1249 | #endif | |
1250 | ||
ac27a0ec | 1251 | #ifdef CONFIG_QUOTA |
d6006186 | 1252 | static const char * const quotatypes[] = INITQFNAMES; |
689c958c | 1253 | #define QTYPE2NAME(t) (quotatypes[t]) |
ac27a0ec | 1254 | |
617ba13b MC |
1255 | static int ext4_write_dquot(struct dquot *dquot); |
1256 | static int ext4_acquire_dquot(struct dquot *dquot); | |
1257 | static int ext4_release_dquot(struct dquot *dquot); | |
1258 | static int ext4_mark_dquot_dirty(struct dquot *dquot); | |
1259 | static int ext4_write_info(struct super_block *sb, int type); | |
6f28e087 | 1260 | static int ext4_quota_on(struct super_block *sb, int type, int format_id, |
8c54ca9c | 1261 | const struct path *path); |
617ba13b MC |
1262 | static int ext4_quota_on_mount(struct super_block *sb, int type); |
1263 | static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, | |
ac27a0ec | 1264 | size_t len, loff_t off); |
617ba13b | 1265 | static ssize_t ext4_quota_write(struct super_block *sb, int type, |
ac27a0ec | 1266 | const char *data, size_t len, loff_t off); |
7c319d32 AK |
1267 | static int ext4_quota_enable(struct super_block *sb, int type, int format_id, |
1268 | unsigned int flags); | |
1269 | static int ext4_enable_quotas(struct super_block *sb); | |
8f0e8746 | 1270 | static int ext4_get_next_id(struct super_block *sb, struct kqid *qid); |
ac27a0ec | 1271 | |
96c7e0d9 JK |
1272 | static struct dquot **ext4_get_dquots(struct inode *inode) |
1273 | { | |
1274 | return EXT4_I(inode)->i_dquot; | |
1275 | } | |
1276 | ||
61e225dc | 1277 | static const struct dquot_operations ext4_quota_operations = { |
7a9ca53a TE |
1278 | .get_reserved_space = ext4_get_reserved_space, |
1279 | .write_dquot = ext4_write_dquot, | |
1280 | .acquire_dquot = ext4_acquire_dquot, | |
1281 | .release_dquot = ext4_release_dquot, | |
1282 | .mark_dirty = ext4_mark_dquot_dirty, | |
1283 | .write_info = ext4_write_info, | |
1284 | .alloc_dquot = dquot_alloc, | |
1285 | .destroy_dquot = dquot_destroy, | |
1286 | .get_projid = ext4_get_projid, | |
1287 | .get_inode_usage = ext4_get_inode_usage, | |
1288 | .get_next_id = ext4_get_next_id, | |
ac27a0ec DK |
1289 | }; |
1290 | ||
0d54b217 | 1291 | static const struct quotactl_ops ext4_qctl_operations = { |
617ba13b | 1292 | .quota_on = ext4_quota_on, |
ca0e05e4 | 1293 | .quota_off = ext4_quota_off, |
287a8095 | 1294 | .quota_sync = dquot_quota_sync, |
0a240339 | 1295 | .get_state = dquot_get_state, |
287a8095 CH |
1296 | .set_info = dquot_set_dqinfo, |
1297 | .get_dqblk = dquot_get_dqblk, | |
6332b9b5 ES |
1298 | .set_dqblk = dquot_set_dqblk, |
1299 | .get_nextdqblk = dquot_get_next_dqblk, | |
ac27a0ec DK |
1300 | }; |
1301 | #endif | |
1302 | ||
ee9b6d61 | 1303 | static const struct super_operations ext4_sops = { |
617ba13b MC |
1304 | .alloc_inode = ext4_alloc_inode, |
1305 | .destroy_inode = ext4_destroy_inode, | |
617ba13b MC |
1306 | .write_inode = ext4_write_inode, |
1307 | .dirty_inode = ext4_dirty_inode, | |
7ff9c073 | 1308 | .drop_inode = ext4_drop_inode, |
0930fcc1 | 1309 | .evict_inode = ext4_evict_inode, |
617ba13b | 1310 | .put_super = ext4_put_super, |
617ba13b | 1311 | .sync_fs = ext4_sync_fs, |
c4be0c1d TS |
1312 | .freeze_fs = ext4_freeze, |
1313 | .unfreeze_fs = ext4_unfreeze, | |
617ba13b MC |
1314 | .statfs = ext4_statfs, |
1315 | .remount_fs = ext4_remount, | |
617ba13b | 1316 | .show_options = ext4_show_options, |
ac27a0ec | 1317 | #ifdef CONFIG_QUOTA |
617ba13b MC |
1318 | .quota_read = ext4_quota_read, |
1319 | .quota_write = ext4_quota_write, | |
96c7e0d9 | 1320 | .get_dquots = ext4_get_dquots, |
ac27a0ec | 1321 | #endif |
c39a7f84 | 1322 | .bdev_try_to_free_page = bdev_try_to_free_page, |
ac27a0ec DK |
1323 | }; |
1324 | ||
39655164 | 1325 | static const struct export_operations ext4_export_ops = { |
1b961ac0 CH |
1326 | .fh_to_dentry = ext4_fh_to_dentry, |
1327 | .fh_to_parent = ext4_fh_to_parent, | |
617ba13b | 1328 | .get_parent = ext4_get_parent, |
ac27a0ec DK |
1329 | }; |
1330 | ||
1331 | enum { | |
1332 | Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, | |
1333 | Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, | |
72578c33 | 1334 | Opt_nouid32, Opt_debug, Opt_removed, |
ac27a0ec | 1335 | Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, |
72578c33 | 1336 | Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, |
ad4eec61 ES |
1337 | Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, |
1338 | Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, | |
ac27a0ec | 1339 | Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, |
6ddb2447 | 1340 | Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, |
ac27a0ec | 1341 | Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, |
5a20bdfc | 1342 | Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, |
ee4a3fcd | 1343 | Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, |
49da9392 | 1344 | Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax, |
1449032b | 1345 | Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit, |
670e9875 | 1346 | Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize, |
1449032b | 1347 | Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, |
5328e635 | 1348 | Opt_inode_readahead_blks, Opt_journal_ioprio, |
744692dc | 1349 | Opt_dioread_nolock, Opt_dioread_lock, |
fc6cb1cd | 1350 | Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, |
cdb7ee4c | 1351 | Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache, |
ac27a0ec DK |
1352 | }; |
1353 | ||
a447c093 | 1354 | static const match_table_t tokens = { |
ac27a0ec DK |
1355 | {Opt_bsd_df, "bsddf"}, |
1356 | {Opt_minix_df, "minixdf"}, | |
1357 | {Opt_grpid, "grpid"}, | |
1358 | {Opt_grpid, "bsdgroups"}, | |
1359 | {Opt_nogrpid, "nogrpid"}, | |
1360 | {Opt_nogrpid, "sysvgroups"}, | |
1361 | {Opt_resgid, "resgid=%u"}, | |
1362 | {Opt_resuid, "resuid=%u"}, | |
1363 | {Opt_sb, "sb=%u"}, | |
1364 | {Opt_err_cont, "errors=continue"}, | |
1365 | {Opt_err_panic, "errors=panic"}, | |
1366 | {Opt_err_ro, "errors=remount-ro"}, | |
1367 | {Opt_nouid32, "nouid32"}, | |
ac27a0ec | 1368 | {Opt_debug, "debug"}, |
72578c33 TT |
1369 | {Opt_removed, "oldalloc"}, |
1370 | {Opt_removed, "orlov"}, | |
ac27a0ec DK |
1371 | {Opt_user_xattr, "user_xattr"}, |
1372 | {Opt_nouser_xattr, "nouser_xattr"}, | |
1373 | {Opt_acl, "acl"}, | |
1374 | {Opt_noacl, "noacl"}, | |
e3bb52ae | 1375 | {Opt_noload, "norecovery"}, |
5a916be1 | 1376 | {Opt_noload, "noload"}, |
72578c33 TT |
1377 | {Opt_removed, "nobh"}, |
1378 | {Opt_removed, "bh"}, | |
ac27a0ec | 1379 | {Opt_commit, "commit=%u"}, |
30773840 TT |
1380 | {Opt_min_batch_time, "min_batch_time=%u"}, |
1381 | {Opt_max_batch_time, "max_batch_time=%u"}, | |
ac27a0ec | 1382 | {Opt_journal_dev, "journal_dev=%u"}, |
ad4eec61 | 1383 | {Opt_journal_path, "journal_path=%s"}, |
818d276c | 1384 | {Opt_journal_checksum, "journal_checksum"}, |
c6d3d56d | 1385 | {Opt_nojournal_checksum, "nojournal_checksum"}, |
818d276c | 1386 | {Opt_journal_async_commit, "journal_async_commit"}, |
ac27a0ec DK |
1387 | {Opt_abort, "abort"}, |
1388 | {Opt_data_journal, "data=journal"}, | |
1389 | {Opt_data_ordered, "data=ordered"}, | |
1390 | {Opt_data_writeback, "data=writeback"}, | |
5bf5683a HK |
1391 | {Opt_data_err_abort, "data_err=abort"}, |
1392 | {Opt_data_err_ignore, "data_err=ignore"}, | |
ac27a0ec DK |
1393 | {Opt_offusrjquota, "usrjquota="}, |
1394 | {Opt_usrjquota, "usrjquota=%s"}, | |
1395 | {Opt_offgrpjquota, "grpjquota="}, | |
1396 | {Opt_grpjquota, "grpjquota=%s"}, | |
1397 | {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, | |
1398 | {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, | |
5a20bdfc | 1399 | {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, |
ac27a0ec DK |
1400 | {Opt_grpquota, "grpquota"}, |
1401 | {Opt_noquota, "noquota"}, | |
1402 | {Opt_quota, "quota"}, | |
1403 | {Opt_usrquota, "usrquota"}, | |
49da9392 | 1404 | {Opt_prjquota, "prjquota"}, |
ac27a0ec | 1405 | {Opt_barrier, "barrier=%u"}, |
06705bff TT |
1406 | {Opt_barrier, "barrier"}, |
1407 | {Opt_nobarrier, "nobarrier"}, | |
25ec56b5 | 1408 | {Opt_i_version, "i_version"}, |
923ae0ff | 1409 | {Opt_dax, "dax"}, |
c9de560d | 1410 | {Opt_stripe, "stripe=%u"}, |
64769240 | 1411 | {Opt_delalloc, "delalloc"}, |
a26f4992 TT |
1412 | {Opt_lazytime, "lazytime"}, |
1413 | {Opt_nolazytime, "nolazytime"}, | |
670e9875 | 1414 | {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"}, |
dd919b98 | 1415 | {Opt_nodelalloc, "nodelalloc"}, |
36ade451 JK |
1416 | {Opt_removed, "mblk_io_submit"}, |
1417 | {Opt_removed, "nomblk_io_submit"}, | |
6fd058f7 TT |
1418 | {Opt_block_validity, "block_validity"}, |
1419 | {Opt_noblock_validity, "noblock_validity"}, | |
240799cd | 1420 | {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, |
b3881f74 | 1421 | {Opt_journal_ioprio, "journal_ioprio=%u"}, |
afd4672d | 1422 | {Opt_auto_da_alloc, "auto_da_alloc=%u"}, |
06705bff TT |
1423 | {Opt_auto_da_alloc, "auto_da_alloc"}, |
1424 | {Opt_noauto_da_alloc, "noauto_da_alloc"}, | |
744692dc JZ |
1425 | {Opt_dioread_nolock, "dioread_nolock"}, |
1426 | {Opt_dioread_lock, "dioread_lock"}, | |
5328e635 ES |
1427 | {Opt_discard, "discard"}, |
1428 | {Opt_nodiscard, "nodiscard"}, | |
fc6cb1cd TT |
1429 | {Opt_init_itable, "init_itable=%u"}, |
1430 | {Opt_init_itable, "init_itable"}, | |
1431 | {Opt_noinit_itable, "noinit_itable"}, | |
df981d03 | 1432 | {Opt_max_dir_size_kb, "max_dir_size_kb=%u"}, |
6ddb2447 | 1433 | {Opt_test_dummy_encryption, "test_dummy_encryption"}, |
cdb7ee4c TE |
1434 | {Opt_nombcache, "nombcache"}, |
1435 | {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */ | |
c7198b9c TT |
1436 | {Opt_removed, "check=none"}, /* mount option from ext2/3 */ |
1437 | {Opt_removed, "nocheck"}, /* mount option from ext2/3 */ | |
1438 | {Opt_removed, "reservation"}, /* mount option from ext2/3 */ | |
1439 | {Opt_removed, "noreservation"}, /* mount option from ext2/3 */ | |
1440 | {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */ | |
f3f12faa | 1441 | {Opt_err, NULL}, |
ac27a0ec DK |
1442 | }; |
1443 | ||
617ba13b | 1444 | static ext4_fsblk_t get_sb_block(void **data) |
ac27a0ec | 1445 | { |
617ba13b | 1446 | ext4_fsblk_t sb_block; |
ac27a0ec DK |
1447 | char *options = (char *) *data; |
1448 | ||
1449 | if (!options || strncmp(options, "sb=", 3) != 0) | |
1450 | return 1; /* Default location */ | |
0b8e58a1 | 1451 | |
ac27a0ec | 1452 | options += 3; |
0b8e58a1 | 1453 | /* TODO: use simple_strtoll with >32bit ext4 */ |
ac27a0ec DK |
1454 | sb_block = simple_strtoul(options, &options, 0); |
1455 | if (*options && *options != ',') { | |
4776004f | 1456 | printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", |
ac27a0ec DK |
1457 | (char *) *data); |
1458 | return 1; | |
1459 | } | |
1460 | if (*options == ',') | |
1461 | options++; | |
1462 | *data = (void *) options; | |
0b8e58a1 | 1463 | |
ac27a0ec DK |
1464 | return sb_block; |
1465 | } | |
1466 | ||
b3881f74 | 1467 | #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) |
d6006186 EB |
1468 | static const char deprecated_msg[] = |
1469 | "Mount option \"%s\" will be removed by %s\n" | |
437ca0fd | 1470 | "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; |
b3881f74 | 1471 | |
56c50f11 DM |
1472 | #ifdef CONFIG_QUOTA |
1473 | static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) | |
1474 | { | |
1475 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1476 | char *qname; | |
03dafb5f | 1477 | int ret = -1; |
56c50f11 DM |
1478 | |
1479 | if (sb_any_quota_loaded(sb) && | |
1480 | !sbi->s_qf_names[qtype]) { | |
1481 | ext4_msg(sb, KERN_ERR, | |
1482 | "Cannot change journaled " | |
1483 | "quota options when quota turned on"); | |
57f73c2c | 1484 | return -1; |
56c50f11 | 1485 | } |
e2b911c5 | 1486 | if (ext4_has_feature_quota(sb)) { |
c325a67c TT |
1487 | ext4_msg(sb, KERN_INFO, "Journaled quota options " |
1488 | "ignored when QUOTA feature is enabled"); | |
1489 | return 1; | |
262b4662 | 1490 | } |
56c50f11 DM |
1491 | qname = match_strdup(args); |
1492 | if (!qname) { | |
1493 | ext4_msg(sb, KERN_ERR, | |
1494 | "Not enough memory for storing quotafile name"); | |
57f73c2c | 1495 | return -1; |
56c50f11 | 1496 | } |
03dafb5f CG |
1497 | if (sbi->s_qf_names[qtype]) { |
1498 | if (strcmp(sbi->s_qf_names[qtype], qname) == 0) | |
1499 | ret = 1; | |
1500 | else | |
1501 | ext4_msg(sb, KERN_ERR, | |
1502 | "%s quota file already specified", | |
1503 | QTYPE2NAME(qtype)); | |
1504 | goto errout; | |
56c50f11 | 1505 | } |
03dafb5f | 1506 | if (strchr(qname, '/')) { |
56c50f11 DM |
1507 | ext4_msg(sb, KERN_ERR, |
1508 | "quotafile must be on filesystem root"); | |
03dafb5f | 1509 | goto errout; |
56c50f11 | 1510 | } |
03dafb5f | 1511 | sbi->s_qf_names[qtype] = qname; |
fd8c37ec | 1512 | set_opt(sb, QUOTA); |
56c50f11 | 1513 | return 1; |
03dafb5f CG |
1514 | errout: |
1515 | kfree(qname); | |
1516 | return ret; | |
56c50f11 DM |
1517 | } |
1518 | ||
1519 | static int clear_qf_name(struct super_block *sb, int qtype) | |
1520 | { | |
1521 | ||
1522 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1523 | ||
1524 | if (sb_any_quota_loaded(sb) && | |
1525 | sbi->s_qf_names[qtype]) { | |
1526 | ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" | |
1527 | " when quota turned on"); | |
57f73c2c | 1528 | return -1; |
56c50f11 | 1529 | } |
03dafb5f | 1530 | kfree(sbi->s_qf_names[qtype]); |
56c50f11 DM |
1531 | sbi->s_qf_names[qtype] = NULL; |
1532 | return 1; | |
1533 | } | |
1534 | #endif | |
1535 | ||
26092bf5 TT |
1536 | #define MOPT_SET 0x0001 |
1537 | #define MOPT_CLEAR 0x0002 | |
1538 | #define MOPT_NOSUPPORT 0x0004 | |
1539 | #define MOPT_EXPLICIT 0x0008 | |
1540 | #define MOPT_CLEAR_ERR 0x0010 | |
1541 | #define MOPT_GTE0 0x0020 | |
ac27a0ec | 1542 | #ifdef CONFIG_QUOTA |
26092bf5 TT |
1543 | #define MOPT_Q 0 |
1544 | #define MOPT_QFMT 0x0040 | |
1545 | #else | |
1546 | #define MOPT_Q MOPT_NOSUPPORT | |
1547 | #define MOPT_QFMT MOPT_NOSUPPORT | |
ac27a0ec | 1548 | #endif |
26092bf5 | 1549 | #define MOPT_DATAJ 0x0080 |
8dc0aa8c TT |
1550 | #define MOPT_NO_EXT2 0x0100 |
1551 | #define MOPT_NO_EXT3 0x0200 | |
1552 | #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) | |
ad4eec61 | 1553 | #define MOPT_STRING 0x0400 |
26092bf5 TT |
1554 | |
1555 | static const struct mount_opts { | |
1556 | int token; | |
1557 | int mount_opt; | |
1558 | int flags; | |
1559 | } ext4_mount_opts[] = { | |
1560 | {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, | |
1561 | {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, | |
1562 | {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, | |
1563 | {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, | |
26092bf5 TT |
1564 | {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, |
1565 | {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, | |
8dc0aa8c TT |
1566 | {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, |
1567 | MOPT_EXT4_ONLY | MOPT_SET}, | |
1568 | {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, | |
1569 | MOPT_EXT4_ONLY | MOPT_CLEAR}, | |
26092bf5 TT |
1570 | {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, |
1571 | {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, | |
8dc0aa8c TT |
1572 | {Opt_delalloc, EXT4_MOUNT_DELALLOC, |
1573 | MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, | |
1574 | {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, | |
59d9fa5c | 1575 | MOPT_EXT4_ONLY | MOPT_CLEAR}, |
c6d3d56d DW |
1576 | {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, |
1577 | MOPT_EXT4_ONLY | MOPT_CLEAR}, | |
8dc0aa8c | 1578 | {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, |
1e381f60 | 1579 | MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, |
26092bf5 | 1580 | {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | |
8dc0aa8c | 1581 | EXT4_MOUNT_JOURNAL_CHECKSUM), |
1e381f60 | 1582 | MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, |
8dc0aa8c | 1583 | {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, |
26092bf5 TT |
1584 | {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR}, |
1585 | {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR}, | |
1586 | {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR}, | |
8dc0aa8c | 1587 | {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, |
7915a861 | 1588 | MOPT_NO_EXT2}, |
8dc0aa8c | 1589 | {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, |
7915a861 | 1590 | MOPT_NO_EXT2}, |
26092bf5 TT |
1591 | {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, |
1592 | {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, | |
1593 | {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, | |
1594 | {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, | |
1595 | {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, | |
1596 | {Opt_commit, 0, MOPT_GTE0}, | |
1597 | {Opt_max_batch_time, 0, MOPT_GTE0}, | |
1598 | {Opt_min_batch_time, 0, MOPT_GTE0}, | |
1599 | {Opt_inode_readahead_blks, 0, MOPT_GTE0}, | |
1600 | {Opt_init_itable, 0, MOPT_GTE0}, | |
923ae0ff | 1601 | {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET}, |
26092bf5 | 1602 | {Opt_stripe, 0, MOPT_GTE0}, |
0efb3b23 JK |
1603 | {Opt_resuid, 0, MOPT_GTE0}, |
1604 | {Opt_resgid, 0, MOPT_GTE0}, | |
5ba92bcf CM |
1605 | {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0}, |
1606 | {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING}, | |
1607 | {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0}, | |
8dc0aa8c TT |
1608 | {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, |
1609 | {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, | |
1610 | {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, | |
1611 | MOPT_NO_EXT2 | MOPT_DATAJ}, | |
26092bf5 TT |
1612 | {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, |
1613 | {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, | |
03010a33 | 1614 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
26092bf5 TT |
1615 | {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, |
1616 | {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, | |
ac27a0ec | 1617 | #else |
26092bf5 TT |
1618 | {Opt_acl, 0, MOPT_NOSUPPORT}, |
1619 | {Opt_noacl, 0, MOPT_NOSUPPORT}, | |
ac27a0ec | 1620 | #endif |
26092bf5 TT |
1621 | {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, |
1622 | {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, | |
670e9875 | 1623 | {Opt_debug_want_extra_isize, 0, MOPT_GTE0}, |
26092bf5 TT |
1624 | {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, |
1625 | {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, | |
1626 | MOPT_SET | MOPT_Q}, | |
1627 | {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, | |
1628 | MOPT_SET | MOPT_Q}, | |
49da9392 JK |
1629 | {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA, |
1630 | MOPT_SET | MOPT_Q}, | |
26092bf5 | 1631 | {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | |
49da9392 JK |
1632 | EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), |
1633 | MOPT_CLEAR | MOPT_Q}, | |
26092bf5 TT |
1634 | {Opt_usrjquota, 0, MOPT_Q}, |
1635 | {Opt_grpjquota, 0, MOPT_Q}, | |
1636 | {Opt_offusrjquota, 0, MOPT_Q}, | |
1637 | {Opt_offgrpjquota, 0, MOPT_Q}, | |
1638 | {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, | |
1639 | {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT}, | |
1640 | {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT}, | |
df981d03 | 1641 | {Opt_max_dir_size_kb, 0, MOPT_GTE0}, |
6ddb2447 | 1642 | {Opt_test_dummy_encryption, 0, MOPT_GTE0}, |
cdb7ee4c | 1643 | {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET}, |
26092bf5 TT |
1644 | {Opt_err, 0, 0} |
1645 | }; | |
1646 | ||
1647 | static int handle_mount_opt(struct super_block *sb, char *opt, int token, | |
1648 | substring_t *args, unsigned long *journal_devnum, | |
1649 | unsigned int *journal_ioprio, int is_remount) | |
1650 | { | |
1651 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1652 | const struct mount_opts *m; | |
08cefc7a EB |
1653 | kuid_t uid; |
1654 | kgid_t gid; | |
26092bf5 TT |
1655 | int arg = 0; |
1656 | ||
57f73c2c TT |
1657 | #ifdef CONFIG_QUOTA |
1658 | if (token == Opt_usrjquota) | |
1659 | return set_qf_name(sb, USRQUOTA, &args[0]); | |
1660 | else if (token == Opt_grpjquota) | |
1661 | return set_qf_name(sb, GRPQUOTA, &args[0]); | |
1662 | else if (token == Opt_offusrjquota) | |
1663 | return clear_qf_name(sb, USRQUOTA); | |
1664 | else if (token == Opt_offgrpjquota) | |
1665 | return clear_qf_name(sb, GRPQUOTA); | |
1666 | #endif | |
26092bf5 | 1667 | switch (token) { |
f7048605 TT |
1668 | case Opt_noacl: |
1669 | case Opt_nouser_xattr: | |
1670 | ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5"); | |
1671 | break; | |
26092bf5 TT |
1672 | case Opt_sb: |
1673 | return 1; /* handled by get_sb_block() */ | |
1674 | case Opt_removed: | |
5f3633e3 | 1675 | ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt); |
26092bf5 | 1676 | return 1; |
26092bf5 TT |
1677 | case Opt_abort: |
1678 | sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; | |
1679 | return 1; | |
1680 | case Opt_i_version: | |
1681 | sb->s_flags |= MS_I_VERSION; | |
1682 | return 1; | |
a26f4992 TT |
1683 | case Opt_lazytime: |
1684 | sb->s_flags |= MS_LAZYTIME; | |
1685 | return 1; | |
1686 | case Opt_nolazytime: | |
1687 | sb->s_flags &= ~MS_LAZYTIME; | |
1688 | return 1; | |
26092bf5 TT |
1689 | } |
1690 | ||
5f3633e3 JK |
1691 | for (m = ext4_mount_opts; m->token != Opt_err; m++) |
1692 | if (token == m->token) | |
1693 | break; | |
1694 | ||
1695 | if (m->token == Opt_err) { | |
1696 | ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" " | |
1697 | "or missing value", opt); | |
1698 | return -1; | |
1699 | } | |
1700 | ||
8dc0aa8c TT |
1701 | if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { |
1702 | ext4_msg(sb, KERN_ERR, | |
1703 | "Mount option \"%s\" incompatible with ext2", opt); | |
1704 | return -1; | |
1705 | } | |
1706 | if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { | |
1707 | ext4_msg(sb, KERN_ERR, | |
1708 | "Mount option \"%s\" incompatible with ext3", opt); | |
1709 | return -1; | |
1710 | } | |
1711 | ||
ad4eec61 | 1712 | if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg)) |
5f3633e3 JK |
1713 | return -1; |
1714 | if (args->from && (m->flags & MOPT_GTE0) && (arg < 0)) | |
1715 | return -1; | |
c93cf2d7 DM |
1716 | if (m->flags & MOPT_EXPLICIT) { |
1717 | if (m->mount_opt & EXT4_MOUNT_DELALLOC) { | |
1718 | set_opt2(sb, EXPLICIT_DELALLOC); | |
1e381f60 DM |
1719 | } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { |
1720 | set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM); | |
c93cf2d7 DM |
1721 | } else |
1722 | return -1; | |
1723 | } | |
5f3633e3 JK |
1724 | if (m->flags & MOPT_CLEAR_ERR) |
1725 | clear_opt(sb, ERRORS_MASK); | |
1726 | if (token == Opt_noquota && sb_any_quota_loaded(sb)) { | |
1727 | ext4_msg(sb, KERN_ERR, "Cannot change quota " | |
1728 | "options when quota turned on"); | |
1729 | return -1; | |
1730 | } | |
1731 | ||
1732 | if (m->flags & MOPT_NOSUPPORT) { | |
1733 | ext4_msg(sb, KERN_ERR, "%s option not supported", opt); | |
1734 | } else if (token == Opt_commit) { | |
1735 | if (arg == 0) | |
1736 | arg = JBD2_DEFAULT_MAX_COMMIT_AGE; | |
1737 | sbi->s_commit_interval = HZ * arg; | |
670e9875 TT |
1738 | } else if (token == Opt_debug_want_extra_isize) { |
1739 | sbi->s_want_extra_isize = arg; | |
5f3633e3 | 1740 | } else if (token == Opt_max_batch_time) { |
5f3633e3 JK |
1741 | sbi->s_max_batch_time = arg; |
1742 | } else if (token == Opt_min_batch_time) { | |
1743 | sbi->s_min_batch_time = arg; | |
1744 | } else if (token == Opt_inode_readahead_blks) { | |
e33e60ea JK |
1745 | if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) { |
1746 | ext4_msg(sb, KERN_ERR, | |
1747 | "EXT4-fs: inode_readahead_blks must be " | |
1748 | "0 or a power of 2 smaller than 2^31"); | |
26092bf5 | 1749 | return -1; |
5f3633e3 JK |
1750 | } |
1751 | sbi->s_inode_readahead_blks = arg; | |
1752 | } else if (token == Opt_init_itable) { | |
1753 | set_opt(sb, INIT_INODE_TABLE); | |
1754 | if (!args->from) | |
1755 | arg = EXT4_DEF_LI_WAIT_MULT; | |
1756 | sbi->s_li_wait_mult = arg; | |
1757 | } else if (token == Opt_max_dir_size_kb) { | |
1758 | sbi->s_max_dir_size_kb = arg; | |
1759 | } else if (token == Opt_stripe) { | |
1760 | sbi->s_stripe = arg; | |
1761 | } else if (token == Opt_resuid) { | |
1762 | uid = make_kuid(current_user_ns(), arg); | |
1763 | if (!uid_valid(uid)) { | |
1764 | ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg); | |
26092bf5 TT |
1765 | return -1; |
1766 | } | |
5f3633e3 JK |
1767 | sbi->s_resuid = uid; |
1768 | } else if (token == Opt_resgid) { | |
1769 | gid = make_kgid(current_user_ns(), arg); | |
1770 | if (!gid_valid(gid)) { | |
1771 | ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg); | |
1772 | return -1; | |
1773 | } | |
1774 | sbi->s_resgid = gid; | |
1775 | } else if (token == Opt_journal_dev) { | |
1776 | if (is_remount) { | |
1777 | ext4_msg(sb, KERN_ERR, | |
1778 | "Cannot specify journal on remount"); | |
1779 | return -1; | |
1780 | } | |
1781 | *journal_devnum = arg; | |
ad4eec61 ES |
1782 | } else if (token == Opt_journal_path) { |
1783 | char *journal_path; | |
1784 | struct inode *journal_inode; | |
1785 | struct path path; | |
1786 | int error; | |
1787 | ||
1788 | if (is_remount) { | |
1789 | ext4_msg(sb, KERN_ERR, | |
1790 | "Cannot specify journal on remount"); | |
1791 | return -1; | |
1792 | } | |
1793 | journal_path = match_strdup(&args[0]); | |
1794 | if (!journal_path) { | |
1795 | ext4_msg(sb, KERN_ERR, "error: could not dup " | |
1796 | "journal device string"); | |
1797 | return -1; | |
1798 | } | |
1799 | ||
1800 | error = kern_path(journal_path, LOOKUP_FOLLOW, &path); | |
1801 | if (error) { | |
1802 | ext4_msg(sb, KERN_ERR, "error: could not find " | |
1803 | "journal device path: error %d", error); | |
1804 | kfree(journal_path); | |
1805 | return -1; | |
1806 | } | |
1807 | ||
2b0143b5 | 1808 | journal_inode = d_inode(path.dentry); |
ad4eec61 ES |
1809 | if (!S_ISBLK(journal_inode->i_mode)) { |
1810 | ext4_msg(sb, KERN_ERR, "error: journal path %s " | |
1811 | "is not a block device", journal_path); | |
1812 | path_put(&path); | |
1813 | kfree(journal_path); | |
1814 | return -1; | |
1815 | } | |
1816 | ||
1817 | *journal_devnum = new_encode_dev(journal_inode->i_rdev); | |
1818 | path_put(&path); | |
1819 | kfree(journal_path); | |
5f3633e3 JK |
1820 | } else if (token == Opt_journal_ioprio) { |
1821 | if (arg > 7) { | |
1822 | ext4_msg(sb, KERN_ERR, "Invalid journal IO priority" | |
1823 | " (must be 0-7)"); | |
1824 | return -1; | |
1825 | } | |
1826 | *journal_ioprio = | |
1827 | IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg); | |
6ddb2447 TT |
1828 | } else if (token == Opt_test_dummy_encryption) { |
1829 | #ifdef CONFIG_EXT4_FS_ENCRYPTION | |
1830 | sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION; | |
1831 | ext4_msg(sb, KERN_WARNING, | |
1832 | "Test dummy encryption mode enabled"); | |
1833 | #else | |
1834 | ext4_msg(sb, KERN_WARNING, | |
1835 | "Test dummy encryption mount option ignored"); | |
1836 | #endif | |
5f3633e3 JK |
1837 | } else if (m->flags & MOPT_DATAJ) { |
1838 | if (is_remount) { | |
1839 | if (!sbi->s_journal) | |
1840 | ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option"); | |
1841 | else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) { | |
0efb3b23 | 1842 | ext4_msg(sb, KERN_ERR, |
26092bf5 | 1843 | "Cannot change data mode on remount"); |
26092bf5 | 1844 | return -1; |
ac27a0ec | 1845 | } |
26092bf5 | 1846 | } else { |
5f3633e3 JK |
1847 | clear_opt(sb, DATA_FLAGS); |
1848 | sbi->s_mount_opt |= m->mount_opt; | |
ac27a0ec | 1849 | } |
5f3633e3 JK |
1850 | #ifdef CONFIG_QUOTA |
1851 | } else if (m->flags & MOPT_QFMT) { | |
1852 | if (sb_any_quota_loaded(sb) && | |
1853 | sbi->s_jquota_fmt != m->mount_opt) { | |
1854 | ext4_msg(sb, KERN_ERR, "Cannot change journaled " | |
1855 | "quota options when quota turned on"); | |
1856 | return -1; | |
1857 | } | |
e2b911c5 | 1858 | if (ext4_has_feature_quota(sb)) { |
c325a67c TT |
1859 | ext4_msg(sb, KERN_INFO, |
1860 | "Quota format mount options ignored " | |
262b4662 | 1861 | "when QUOTA feature is enabled"); |
c325a67c | 1862 | return 1; |
262b4662 | 1863 | } |
5f3633e3 | 1864 | sbi->s_jquota_fmt = m->mount_opt; |
923ae0ff | 1865 | #endif |
923ae0ff | 1866 | } else if (token == Opt_dax) { |
ef83b6e8 DW |
1867 | #ifdef CONFIG_FS_DAX |
1868 | ext4_msg(sb, KERN_WARNING, | |
1869 | "DAX enabled. Warning: EXPERIMENTAL, use at your own risk"); | |
1870 | sbi->s_mount_opt |= m->mount_opt; | |
1871 | #else | |
923ae0ff RZ |
1872 | ext4_msg(sb, KERN_INFO, "dax option not supported"); |
1873 | return -1; | |
5f3633e3 | 1874 | #endif |
7915a861 AN |
1875 | } else if (token == Opt_data_err_abort) { |
1876 | sbi->s_mount_opt |= m->mount_opt; | |
1877 | } else if (token == Opt_data_err_ignore) { | |
1878 | sbi->s_mount_opt &= ~m->mount_opt; | |
5f3633e3 JK |
1879 | } else { |
1880 | if (!args->from) | |
1881 | arg = 1; | |
1882 | if (m->flags & MOPT_CLEAR) | |
1883 | arg = !arg; | |
1884 | else if (unlikely(!(m->flags & MOPT_SET))) { | |
1885 | ext4_msg(sb, KERN_WARNING, | |
1886 | "buggy handling of option %s", opt); | |
1887 | WARN_ON(1); | |
1888 | return -1; | |
1889 | } | |
1890 | if (arg != 0) | |
1891 | sbi->s_mount_opt |= m->mount_opt; | |
1892 | else | |
1893 | sbi->s_mount_opt &= ~m->mount_opt; | |
26092bf5 | 1894 | } |
5f3633e3 | 1895 | return 1; |
26092bf5 TT |
1896 | } |
1897 | ||
1898 | static int parse_options(char *options, struct super_block *sb, | |
1899 | unsigned long *journal_devnum, | |
1900 | unsigned int *journal_ioprio, | |
1901 | int is_remount) | |
1902 | { | |
1903 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1904 | char *p; | |
1905 | substring_t args[MAX_OPT_ARGS]; | |
1906 | int token; | |
1907 | ||
1908 | if (!options) | |
1909 | return 1; | |
1910 | ||
1911 | while ((p = strsep(&options, ",")) != NULL) { | |
1912 | if (!*p) | |
1913 | continue; | |
1914 | /* | |
1915 | * Initialize args struct so we know whether arg was | |
1916 | * found; some options take optional arguments. | |
1917 | */ | |
caecd0af | 1918 | args[0].to = args[0].from = NULL; |
26092bf5 TT |
1919 | token = match_token(p, tokens, args); |
1920 | if (handle_mount_opt(sb, p, token, args, journal_devnum, | |
1921 | journal_ioprio, is_remount) < 0) | |
1922 | return 0; | |
ac27a0ec DK |
1923 | } |
1924 | #ifdef CONFIG_QUOTA | |
49da9392 JK |
1925 | /* |
1926 | * We do the test below only for project quotas. 'usrquota' and | |
1927 | * 'grpquota' mount options are allowed even without quota feature | |
1928 | * to support legacy quotas in quota files. | |
1929 | */ | |
1930 | if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) { | |
1931 | ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. " | |
1932 | "Cannot enable project quota enforcement."); | |
1933 | return 0; | |
1934 | } | |
1935 | if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { | |
482a7425 | 1936 | if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA]) |
fd8c37ec | 1937 | clear_opt(sb, USRQUOTA); |
ac27a0ec | 1938 | |
482a7425 | 1939 | if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA]) |
fd8c37ec | 1940 | clear_opt(sb, GRPQUOTA); |
ac27a0ec | 1941 | |
56c50f11 | 1942 | if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { |
b31e1552 ES |
1943 | ext4_msg(sb, KERN_ERR, "old and new quota " |
1944 | "format mixing"); | |
ac27a0ec DK |
1945 | return 0; |
1946 | } | |
1947 | ||
1948 | if (!sbi->s_jquota_fmt) { | |
b31e1552 ES |
1949 | ext4_msg(sb, KERN_ERR, "journaled quota format " |
1950 | "not specified"); | |
ac27a0ec DK |
1951 | return 0; |
1952 | } | |
ac27a0ec DK |
1953 | } |
1954 | #endif | |
261cb20c JK |
1955 | if (test_opt(sb, DIOREAD_NOLOCK)) { |
1956 | int blocksize = | |
1957 | BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); | |
1958 | ||
09cbfeaf | 1959 | if (blocksize < PAGE_SIZE) { |
261cb20c JK |
1960 | ext4_msg(sb, KERN_ERR, "can't mount with " |
1961 | "dioread_nolock if block size != PAGE_SIZE"); | |
1962 | return 0; | |
1963 | } | |
1964 | } | |
ac27a0ec DK |
1965 | return 1; |
1966 | } | |
1967 | ||
2adf6da8 TT |
1968 | static inline void ext4_show_quota_options(struct seq_file *seq, |
1969 | struct super_block *sb) | |
1970 | { | |
1971 | #if defined(CONFIG_QUOTA) | |
1972 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1973 | ||
1974 | if (sbi->s_jquota_fmt) { | |
1975 | char *fmtname = ""; | |
1976 | ||
1977 | switch (sbi->s_jquota_fmt) { | |
1978 | case QFMT_VFS_OLD: | |
1979 | fmtname = "vfsold"; | |
1980 | break; | |
1981 | case QFMT_VFS_V0: | |
1982 | fmtname = "vfsv0"; | |
1983 | break; | |
1984 | case QFMT_VFS_V1: | |
1985 | fmtname = "vfsv1"; | |
1986 | break; | |
1987 | } | |
1988 | seq_printf(seq, ",jqfmt=%s", fmtname); | |
1989 | } | |
1990 | ||
1991 | if (sbi->s_qf_names[USRQUOTA]) | |
a068acf2 | 1992 | seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]); |
2adf6da8 TT |
1993 | |
1994 | if (sbi->s_qf_names[GRPQUOTA]) | |
a068acf2 | 1995 | seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]); |
2adf6da8 TT |
1996 | #endif |
1997 | } | |
1998 | ||
5a916be1 TT |
1999 | static const char *token2str(int token) |
2000 | { | |
50df9fd5 | 2001 | const struct match_token *t; |
5a916be1 TT |
2002 | |
2003 | for (t = tokens; t->token != Opt_err; t++) | |
2004 | if (t->token == token && !strchr(t->pattern, '=')) | |
2005 | break; | |
2006 | return t->pattern; | |
2007 | } | |
2008 | ||
2adf6da8 TT |
2009 | /* |
2010 | * Show an option if | |
2011 | * - it's set to a non-default value OR | |
2012 | * - if the per-sb default is different from the global default | |
2013 | */ | |
66acdcf4 TT |
2014 | static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, |
2015 | int nodefs) | |
2adf6da8 | 2016 | { |
2adf6da8 TT |
2017 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
2018 | struct ext4_super_block *es = sbi->s_es; | |
66acdcf4 | 2019 | int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt; |
5a916be1 | 2020 | const struct mount_opts *m; |
66acdcf4 | 2021 | char sep = nodefs ? '\n' : ','; |
2adf6da8 | 2022 | |
66acdcf4 TT |
2023 | #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) |
2024 | #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) | |
2adf6da8 TT |
2025 | |
2026 | if (sbi->s_sb_block != 1) | |
5a916be1 TT |
2027 | SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); |
2028 | ||
2029 | for (m = ext4_mount_opts; m->token != Opt_err; m++) { | |
2030 | int want_set = m->flags & MOPT_SET; | |
2031 | if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || | |
2032 | (m->flags & MOPT_CLEAR_ERR)) | |
2033 | continue; | |
66acdcf4 | 2034 | if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) |
5a916be1 TT |
2035 | continue; /* skip if same as the default */ |
2036 | if ((want_set && | |
2037 | (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || | |
2038 | (!want_set && (sbi->s_mount_opt & m->mount_opt))) | |
2039 | continue; /* select Opt_noFoo vs Opt_Foo */ | |
2040 | SEQ_OPTS_PRINT("%s", token2str(m->token)); | |
2adf6da8 | 2041 | } |
5a916be1 | 2042 | |
08cefc7a | 2043 | if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || |
5a916be1 | 2044 | le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) |
08cefc7a EB |
2045 | SEQ_OPTS_PRINT("resuid=%u", |
2046 | from_kuid_munged(&init_user_ns, sbi->s_resuid)); | |
2047 | if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || | |
5a916be1 | 2048 | le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) |
08cefc7a EB |
2049 | SEQ_OPTS_PRINT("resgid=%u", |
2050 | from_kgid_munged(&init_user_ns, sbi->s_resgid)); | |
66acdcf4 | 2051 | def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); |
5a916be1 TT |
2052 | if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) |
2053 | SEQ_OPTS_PUTS("errors=remount-ro"); | |
2adf6da8 | 2054 | if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) |
5a916be1 | 2055 | SEQ_OPTS_PUTS("errors=continue"); |
2adf6da8 | 2056 | if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) |
5a916be1 | 2057 | SEQ_OPTS_PUTS("errors=panic"); |
66acdcf4 | 2058 | if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) |
5a916be1 | 2059 | SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); |
66acdcf4 | 2060 | if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) |
5a916be1 | 2061 | SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); |
66acdcf4 | 2062 | if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) |
5a916be1 | 2063 | SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); |
2adf6da8 | 2064 | if (sb->s_flags & MS_I_VERSION) |
5a916be1 | 2065 | SEQ_OPTS_PUTS("i_version"); |
66acdcf4 | 2066 | if (nodefs || sbi->s_stripe) |
5a916be1 | 2067 | SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); |
66acdcf4 | 2068 | if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) { |
5a916be1 TT |
2069 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) |
2070 | SEQ_OPTS_PUTS("data=journal"); | |
2071 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) | |
2072 | SEQ_OPTS_PUTS("data=ordered"); | |
2073 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) | |
2074 | SEQ_OPTS_PUTS("data=writeback"); | |
2075 | } | |
66acdcf4 TT |
2076 | if (nodefs || |
2077 | sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) | |
5a916be1 TT |
2078 | SEQ_OPTS_PRINT("inode_readahead_blks=%u", |
2079 | sbi->s_inode_readahead_blks); | |
2adf6da8 | 2080 | |
66acdcf4 TT |
2081 | if (nodefs || (test_opt(sb, INIT_INODE_TABLE) && |
2082 | (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) | |
5a916be1 | 2083 | SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); |
df981d03 TT |
2084 | if (nodefs || sbi->s_max_dir_size_kb) |
2085 | SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); | |
7915a861 AN |
2086 | if (test_opt(sb, DATA_ERR_ABORT)) |
2087 | SEQ_OPTS_PUTS("data_err=abort"); | |
2adf6da8 TT |
2088 | |
2089 | ext4_show_quota_options(seq, sb); | |
2adf6da8 TT |
2090 | return 0; |
2091 | } | |
2092 | ||
66acdcf4 TT |
2093 | static int ext4_show_options(struct seq_file *seq, struct dentry *root) |
2094 | { | |
2095 | return _ext4_show_options(seq, root->d_sb, 0); | |
2096 | } | |
2097 | ||
ebd173be | 2098 | int ext4_seq_options_show(struct seq_file *seq, void *offset) |
66acdcf4 TT |
2099 | { |
2100 | struct super_block *sb = seq->private; | |
2101 | int rc; | |
2102 | ||
bc98a42c | 2103 | seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw"); |
66acdcf4 TT |
2104 | rc = _ext4_show_options(seq, sb, 1); |
2105 | seq_puts(seq, "\n"); | |
2106 | return rc; | |
2107 | } | |
2108 | ||
617ba13b | 2109 | static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, |
ac27a0ec DK |
2110 | int read_only) |
2111 | { | |
617ba13b | 2112 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec DK |
2113 | int res = 0; |
2114 | ||
617ba13b | 2115 | if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { |
b31e1552 ES |
2116 | ext4_msg(sb, KERN_ERR, "revision level too high, " |
2117 | "forcing read-only mode"); | |
ac27a0ec DK |
2118 | res = MS_RDONLY; |
2119 | } | |
2120 | if (read_only) | |
281b5995 | 2121 | goto done; |
617ba13b | 2122 | if (!(sbi->s_mount_state & EXT4_VALID_FS)) |
b31e1552 ES |
2123 | ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " |
2124 | "running e2fsck is recommended"); | |
c8b459f4 | 2125 | else if (sbi->s_mount_state & EXT4_ERROR_FS) |
b31e1552 ES |
2126 | ext4_msg(sb, KERN_WARNING, |
2127 | "warning: mounting fs with errors, " | |
2128 | "running e2fsck is recommended"); | |
ed3ce80a | 2129 | else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && |
ac27a0ec DK |
2130 | le16_to_cpu(es->s_mnt_count) >= |
2131 | (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) | |
b31e1552 ES |
2132 | ext4_msg(sb, KERN_WARNING, |
2133 | "warning: maximal mount count reached, " | |
2134 | "running e2fsck is recommended"); | |
ac27a0ec DK |
2135 | else if (le32_to_cpu(es->s_checkinterval) && |
2136 | (le32_to_cpu(es->s_lastcheck) + | |
2137 | le32_to_cpu(es->s_checkinterval) <= get_seconds())) | |
b31e1552 ES |
2138 | ext4_msg(sb, KERN_WARNING, |
2139 | "warning: checktime reached, " | |
2140 | "running e2fsck is recommended"); | |
0b8e58a1 | 2141 | if (!sbi->s_journal) |
0390131b | 2142 | es->s_state &= cpu_to_le16(~EXT4_VALID_FS); |
ac27a0ec | 2143 | if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) |
617ba13b | 2144 | es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); |
e8546d06 | 2145 | le16_add_cpu(&es->s_mnt_count, 1); |
ac27a0ec | 2146 | es->s_mtime = cpu_to_le32(get_seconds()); |
617ba13b | 2147 | ext4_update_dynamic_rev(sb); |
0390131b | 2148 | if (sbi->s_journal) |
e2b911c5 | 2149 | ext4_set_feature_journal_needs_recovery(sb); |
ac27a0ec | 2150 | |
e2d67052 | 2151 | ext4_commit_super(sb, 1); |
281b5995 | 2152 | done: |
ac27a0ec | 2153 | if (test_opt(sb, DEBUG)) |
a9df9a49 | 2154 | printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " |
a2595b8a | 2155 | "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", |
ac27a0ec DK |
2156 | sb->s_blocksize, |
2157 | sbi->s_groups_count, | |
617ba13b MC |
2158 | EXT4_BLOCKS_PER_GROUP(sb), |
2159 | EXT4_INODES_PER_GROUP(sb), | |
a2595b8a | 2160 | sbi->s_mount_opt, sbi->s_mount_opt2); |
ac27a0ec | 2161 | |
7abc52c2 | 2162 | cleancache_init_fs(sb); |
ac27a0ec DK |
2163 | return res; |
2164 | } | |
2165 | ||
117fff10 TT |
2166 | int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) |
2167 | { | |
2168 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
2169 | struct flex_groups *new_groups; | |
2170 | int size; | |
2171 | ||
2172 | if (!sbi->s_log_groups_per_flex) | |
2173 | return 0; | |
2174 | ||
2175 | size = ext4_flex_group(sbi, ngroup - 1) + 1; | |
2176 | if (size <= sbi->s_flex_groups_allocated) | |
2177 | return 0; | |
2178 | ||
2179 | size = roundup_pow_of_two(size * sizeof(struct flex_groups)); | |
a7c3e901 | 2180 | new_groups = kvzalloc(size, GFP_KERNEL); |
117fff10 TT |
2181 | if (!new_groups) { |
2182 | ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups", | |
2183 | size / (int) sizeof(struct flex_groups)); | |
2184 | return -ENOMEM; | |
2185 | } | |
2186 | ||
2187 | if (sbi->s_flex_groups) { | |
2188 | memcpy(new_groups, sbi->s_flex_groups, | |
2189 | (sbi->s_flex_groups_allocated * | |
2190 | sizeof(struct flex_groups))); | |
b93b41d4 | 2191 | kvfree(sbi->s_flex_groups); |
117fff10 TT |
2192 | } |
2193 | sbi->s_flex_groups = new_groups; | |
2194 | sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups); | |
2195 | return 0; | |
2196 | } | |
2197 | ||
772cb7c8 JS |
2198 | static int ext4_fill_flex_info(struct super_block *sb) |
2199 | { | |
2200 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
2201 | struct ext4_group_desc *gdp = NULL; | |
772cb7c8 | 2202 | ext4_group_t flex_group; |
117fff10 | 2203 | int i, err; |
772cb7c8 | 2204 | |
503358ae | 2205 | sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; |
d50f2ab6 | 2206 | if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { |
772cb7c8 JS |
2207 | sbi->s_log_groups_per_flex = 0; |
2208 | return 1; | |
2209 | } | |
2210 | ||
117fff10 TT |
2211 | err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); |
2212 | if (err) | |
9933fc0a | 2213 | goto failed; |
772cb7c8 | 2214 | |
772cb7c8 | 2215 | for (i = 0; i < sbi->s_groups_count; i++) { |
88b6edd1 | 2216 | gdp = ext4_get_group_desc(sb, i, NULL); |
772cb7c8 JS |
2217 | |
2218 | flex_group = ext4_flex_group(sbi, i); | |
7ad9bb65 TT |
2219 | atomic_add(ext4_free_inodes_count(sb, gdp), |
2220 | &sbi->s_flex_groups[flex_group].free_inodes); | |
90ba983f TT |
2221 | atomic64_add(ext4_free_group_clusters(sb, gdp), |
2222 | &sbi->s_flex_groups[flex_group].free_clusters); | |
7ad9bb65 TT |
2223 | atomic_add(ext4_used_dirs_count(sb, gdp), |
2224 | &sbi->s_flex_groups[flex_group].used_dirs); | |
772cb7c8 JS |
2225 | } |
2226 | ||
2227 | return 1; | |
2228 | failed: | |
2229 | return 0; | |
2230 | } | |
2231 | ||
e2b911c5 | 2232 | static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, |
feb0ab32 | 2233 | struct ext4_group_desc *gdp) |
717d50e4 | 2234 | { |
b47820ed | 2235 | int offset = offsetof(struct ext4_group_desc, bg_checksum); |
717d50e4 | 2236 | __u16 crc = 0; |
feb0ab32 | 2237 | __le32 le_group = cpu_to_le32(block_group); |
e2b911c5 | 2238 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
717d50e4 | 2239 | |
9aa5d32b | 2240 | if (ext4_has_metadata_csum(sbi->s_sb)) { |
feb0ab32 | 2241 | /* Use new metadata_csum algorithm */ |
feb0ab32 | 2242 | __u32 csum32; |
b47820ed | 2243 | __u16 dummy_csum = 0; |
feb0ab32 | 2244 | |
feb0ab32 DW |
2245 | csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, |
2246 | sizeof(le_group)); | |
b47820ed DJ |
2247 | csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset); |
2248 | csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum, | |
2249 | sizeof(dummy_csum)); | |
2250 | offset += sizeof(dummy_csum); | |
2251 | if (offset < sbi->s_desc_size) | |
2252 | csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset, | |
2253 | sbi->s_desc_size - offset); | |
feb0ab32 DW |
2254 | |
2255 | crc = csum32 & 0xFFFF; | |
2256 | goto out; | |
717d50e4 AD |
2257 | } |
2258 | ||
feb0ab32 | 2259 | /* old crc16 code */ |
e2b911c5 | 2260 | if (!ext4_has_feature_gdt_csum(sb)) |
813d32f9 DW |
2261 | return 0; |
2262 | ||
feb0ab32 DW |
2263 | crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); |
2264 | crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); | |
2265 | crc = crc16(crc, (__u8 *)gdp, offset); | |
2266 | offset += sizeof(gdp->bg_checksum); /* skip checksum */ | |
2267 | /* for checksum of struct ext4_group_desc do the rest...*/ | |
e2b911c5 | 2268 | if (ext4_has_feature_64bit(sb) && |
feb0ab32 DW |
2269 | offset < le16_to_cpu(sbi->s_es->s_desc_size)) |
2270 | crc = crc16(crc, (__u8 *)gdp + offset, | |
2271 | le16_to_cpu(sbi->s_es->s_desc_size) - | |
2272 | offset); | |
2273 | ||
2274 | out: | |
717d50e4 AD |
2275 | return cpu_to_le16(crc); |
2276 | } | |
2277 | ||
feb0ab32 | 2278 | int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, |
717d50e4 AD |
2279 | struct ext4_group_desc *gdp) |
2280 | { | |
feb0ab32 | 2281 | if (ext4_has_group_desc_csum(sb) && |
e2b911c5 | 2282 | (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) |
717d50e4 AD |
2283 | return 0; |
2284 | ||
2285 | return 1; | |
2286 | } | |
2287 | ||
feb0ab32 DW |
2288 | void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, |
2289 | struct ext4_group_desc *gdp) | |
2290 | { | |
2291 | if (!ext4_has_group_desc_csum(sb)) | |
2292 | return; | |
e2b911c5 | 2293 | gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); |
feb0ab32 DW |
2294 | } |
2295 | ||
ac27a0ec | 2296 | /* Called at mount-time, super-block is locked */ |
bfff6873 | 2297 | static int ext4_check_descriptors(struct super_block *sb, |
829fa70d | 2298 | ext4_fsblk_t sb_block, |
bfff6873 | 2299 | ext4_group_t *first_not_zeroed) |
ac27a0ec | 2300 | { |
617ba13b MC |
2301 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
2302 | ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); | |
2303 | ext4_fsblk_t last_block; | |
bd81d8ee LV |
2304 | ext4_fsblk_t block_bitmap; |
2305 | ext4_fsblk_t inode_bitmap; | |
2306 | ext4_fsblk_t inode_table; | |
ce421581 | 2307 | int flexbg_flag = 0; |
bfff6873 | 2308 | ext4_group_t i, grp = sbi->s_groups_count; |
ac27a0ec | 2309 | |
e2b911c5 | 2310 | if (ext4_has_feature_flex_bg(sb)) |
ce421581 JS |
2311 | flexbg_flag = 1; |
2312 | ||
af5bc92d | 2313 | ext4_debug("Checking group descriptors"); |
ac27a0ec | 2314 | |
197cd65a AM |
2315 | for (i = 0; i < sbi->s_groups_count; i++) { |
2316 | struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); | |
2317 | ||
ce421581 | 2318 | if (i == sbi->s_groups_count - 1 || flexbg_flag) |
bd81d8ee | 2319 | last_block = ext4_blocks_count(sbi->s_es) - 1; |
ac27a0ec DK |
2320 | else |
2321 | last_block = first_block + | |
617ba13b | 2322 | (EXT4_BLOCKS_PER_GROUP(sb) - 1); |
ac27a0ec | 2323 | |
bfff6873 LC |
2324 | if ((grp == sbi->s_groups_count) && |
2325 | !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) | |
2326 | grp = i; | |
2327 | ||
8fadc143 | 2328 | block_bitmap = ext4_block_bitmap(sb, gdp); |
829fa70d TT |
2329 | if (block_bitmap == sb_block) { |
2330 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " | |
2331 | "Block bitmap for group %u overlaps " | |
2332 | "superblock", i); | |
2333 | } | |
2b2d6d01 | 2334 | if (block_bitmap < first_block || block_bitmap > last_block) { |
b31e1552 | 2335 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 2336 | "Block bitmap for group %u not in group " |
b31e1552 | 2337 | "(block %llu)!", i, block_bitmap); |
ac27a0ec DK |
2338 | return 0; |
2339 | } | |
8fadc143 | 2340 | inode_bitmap = ext4_inode_bitmap(sb, gdp); |
829fa70d TT |
2341 | if (inode_bitmap == sb_block) { |
2342 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " | |
2343 | "Inode bitmap for group %u overlaps " | |
2344 | "superblock", i); | |
2345 | } | |
2b2d6d01 | 2346 | if (inode_bitmap < first_block || inode_bitmap > last_block) { |
b31e1552 | 2347 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 2348 | "Inode bitmap for group %u not in group " |
b31e1552 | 2349 | "(block %llu)!", i, inode_bitmap); |
ac27a0ec DK |
2350 | return 0; |
2351 | } | |
8fadc143 | 2352 | inode_table = ext4_inode_table(sb, gdp); |
829fa70d TT |
2353 | if (inode_table == sb_block) { |
2354 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " | |
2355 | "Inode table for group %u overlaps " | |
2356 | "superblock", i); | |
2357 | } | |
bd81d8ee | 2358 | if (inode_table < first_block || |
2b2d6d01 | 2359 | inode_table + sbi->s_itb_per_group - 1 > last_block) { |
b31e1552 | 2360 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 2361 | "Inode table for group %u not in group " |
b31e1552 | 2362 | "(block %llu)!", i, inode_table); |
ac27a0ec DK |
2363 | return 0; |
2364 | } | |
955ce5f5 | 2365 | ext4_lock_group(sb, i); |
feb0ab32 | 2366 | if (!ext4_group_desc_csum_verify(sb, i, gdp)) { |
b31e1552 ES |
2367 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
2368 | "Checksum for group %u failed (%u!=%u)", | |
e2b911c5 | 2369 | i, le16_to_cpu(ext4_group_desc_csum(sb, i, |
b31e1552 | 2370 | gdp)), le16_to_cpu(gdp->bg_checksum)); |
bc98a42c | 2371 | if (!sb_rdonly(sb)) { |
955ce5f5 | 2372 | ext4_unlock_group(sb, i); |
8a266467 | 2373 | return 0; |
7ee1ec4c | 2374 | } |
717d50e4 | 2375 | } |
955ce5f5 | 2376 | ext4_unlock_group(sb, i); |
ce421581 JS |
2377 | if (!flexbg_flag) |
2378 | first_block += EXT4_BLOCKS_PER_GROUP(sb); | |
ac27a0ec | 2379 | } |
bfff6873 LC |
2380 | if (NULL != first_not_zeroed) |
2381 | *first_not_zeroed = grp; | |
ac27a0ec DK |
2382 | return 1; |
2383 | } | |
2384 | ||
617ba13b | 2385 | /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at |
ac27a0ec DK |
2386 | * the superblock) which were deleted from all directories, but held open by |
2387 | * a process at the time of a crash. We walk the list and try to delete these | |
2388 | * inodes at recovery time (only with a read-write filesystem). | |
2389 | * | |
2390 | * In order to keep the orphan inode chain consistent during traversal (in | |
2391 | * case of crash during recovery), we link each inode into the superblock | |
2392 | * orphan list_head and handle it the same way as an inode deletion during | |
2393 | * normal operation (which journals the operations for us). | |
2394 | * | |
2395 | * We only do an iget() and an iput() on each inode, which is very safe if we | |
2396 | * accidentally point at an in-use or already deleted inode. The worst that | |
2397 | * can happen in this case is that we get a "bit already cleared" message from | |
617ba13b | 2398 | * ext4_free_inode(). The only reason we would point at a wrong inode is if |
ac27a0ec DK |
2399 | * e2fsck was run on this filesystem, and it must have already done the orphan |
2400 | * inode cleanup for us, so we can safely abort without any further action. | |
2401 | */ | |
2b2d6d01 TT |
2402 | static void ext4_orphan_cleanup(struct super_block *sb, |
2403 | struct ext4_super_block *es) | |
ac27a0ec DK |
2404 | { |
2405 | unsigned int s_flags = sb->s_flags; | |
2c98eb5e | 2406 | int ret, nr_orphans = 0, nr_truncates = 0; |
ac27a0ec DK |
2407 | #ifdef CONFIG_QUOTA |
2408 | int i; | |
2409 | #endif | |
2410 | if (!es->s_last_orphan) { | |
2411 | jbd_debug(4, "no orphan inodes to clean up\n"); | |
2412 | return; | |
2413 | } | |
2414 | ||
a8f48a95 | 2415 | if (bdev_read_only(sb->s_bdev)) { |
b31e1552 ES |
2416 | ext4_msg(sb, KERN_ERR, "write access " |
2417 | "unavailable, skipping orphan cleanup"); | |
a8f48a95 ES |
2418 | return; |
2419 | } | |
2420 | ||
d39195c3 AG |
2421 | /* Check if feature set would not allow a r/w mount */ |
2422 | if (!ext4_feature_set_ok(sb, 0)) { | |
2423 | ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " | |
2424 | "unknown ROCOMPAT features"); | |
2425 | return; | |
2426 | } | |
2427 | ||
617ba13b | 2428 | if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { |
c25f9bc6 ES |
2429 | /* don't clear list on RO mount w/ errors */ |
2430 | if (es->s_last_orphan && !(s_flags & MS_RDONLY)) { | |
84474976 | 2431 | ext4_msg(sb, KERN_INFO, "Errors on filesystem, " |
ac27a0ec | 2432 | "clearing orphan list.\n"); |
c25f9bc6 ES |
2433 | es->s_last_orphan = 0; |
2434 | } | |
ac27a0ec DK |
2435 | jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); |
2436 | return; | |
2437 | } | |
2438 | ||
2439 | if (s_flags & MS_RDONLY) { | |
b31e1552 | 2440 | ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); |
ac27a0ec DK |
2441 | sb->s_flags &= ~MS_RDONLY; |
2442 | } | |
2443 | #ifdef CONFIG_QUOTA | |
2444 | /* Needed for iput() to work correctly and not trash data */ | |
2445 | sb->s_flags |= MS_ACTIVE; | |
2446 | /* Turn on quotas so that they are updated correctly */ | |
a2d4a646 | 2447 | for (i = 0; i < EXT4_MAXQUOTAS; i++) { |
617ba13b MC |
2448 | if (EXT4_SB(sb)->s_qf_names[i]) { |
2449 | int ret = ext4_quota_on_mount(sb, i); | |
ac27a0ec | 2450 | if (ret < 0) |
b31e1552 ES |
2451 | ext4_msg(sb, KERN_ERR, |
2452 | "Cannot turn on journaled " | |
2453 | "quota: error %d", ret); | |
ac27a0ec DK |
2454 | } |
2455 | } | |
2456 | #endif | |
2457 | ||
2458 | while (es->s_last_orphan) { | |
2459 | struct inode *inode; | |
2460 | ||
c65d5c6c VN |
2461 | /* |
2462 | * We may have encountered an error during cleanup; if | |
2463 | * so, skip the rest. | |
2464 | */ | |
2465 | if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { | |
2466 | jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); | |
2467 | es->s_last_orphan = 0; | |
2468 | break; | |
2469 | } | |
2470 | ||
97bd42b9 JB |
2471 | inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); |
2472 | if (IS_ERR(inode)) { | |
ac27a0ec DK |
2473 | es->s_last_orphan = 0; |
2474 | break; | |
2475 | } | |
2476 | ||
617ba13b | 2477 | list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); |
871a2931 | 2478 | dquot_initialize(inode); |
ac27a0ec | 2479 | if (inode->i_nlink) { |
566370a2 PT |
2480 | if (test_opt(sb, DEBUG)) |
2481 | ext4_msg(sb, KERN_DEBUG, | |
2482 | "%s: truncating inode %lu to %lld bytes", | |
2483 | __func__, inode->i_ino, inode->i_size); | |
e5f8eab8 | 2484 | jbd_debug(2, "truncating inode %lu to %lld bytes\n", |
ac27a0ec | 2485 | inode->i_ino, inode->i_size); |
5955102c | 2486 | inode_lock(inode); |
55f252c9 | 2487 | truncate_inode_pages(inode->i_mapping, inode->i_size); |
2c98eb5e TT |
2488 | ret = ext4_truncate(inode); |
2489 | if (ret) | |
2490 | ext4_std_error(inode->i_sb, ret); | |
5955102c | 2491 | inode_unlock(inode); |
ac27a0ec DK |
2492 | nr_truncates++; |
2493 | } else { | |
566370a2 PT |
2494 | if (test_opt(sb, DEBUG)) |
2495 | ext4_msg(sb, KERN_DEBUG, | |
2496 | "%s: deleting unreferenced inode %lu", | |
2497 | __func__, inode->i_ino); | |
ac27a0ec DK |
2498 | jbd_debug(2, "deleting unreferenced inode %lu\n", |
2499 | inode->i_ino); | |
2500 | nr_orphans++; | |
2501 | } | |
2502 | iput(inode); /* The delete magic happens here! */ | |
2503 | } | |
2504 | ||
2b2d6d01 | 2505 | #define PLURAL(x) (x), ((x) == 1) ? "" : "s" |
ac27a0ec DK |
2506 | |
2507 | if (nr_orphans) | |
b31e1552 ES |
2508 | ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", |
2509 | PLURAL(nr_orphans)); | |
ac27a0ec | 2510 | if (nr_truncates) |
b31e1552 ES |
2511 | ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", |
2512 | PLURAL(nr_truncates)); | |
ac27a0ec DK |
2513 | #ifdef CONFIG_QUOTA |
2514 | /* Turn quotas off */ | |
a2d4a646 | 2515 | for (i = 0; i < EXT4_MAXQUOTAS; i++) { |
ac27a0ec | 2516 | if (sb_dqopt(sb)->files[i]) |
287a8095 | 2517 | dquot_quota_off(sb, i); |
ac27a0ec DK |
2518 | } |
2519 | #endif | |
2520 | sb->s_flags = s_flags; /* Restore MS_RDONLY status */ | |
2521 | } | |
0b8e58a1 | 2522 | |
cd2291a4 ES |
2523 | /* |
2524 | * Maximal extent format file size. | |
2525 | * Resulting logical blkno at s_maxbytes must fit in our on-disk | |
2526 | * extent format containers, within a sector_t, and within i_blocks | |
2527 | * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, | |
2528 | * so that won't be a limiting factor. | |
2529 | * | |
f17722f9 LC |
2530 | * However there is other limiting factor. We do store extents in the form |
2531 | * of starting block and length, hence the resulting length of the extent | |
2532 | * covering maximum file size must fit into on-disk format containers as | |
2533 | * well. Given that length is always by 1 unit bigger than max unit (because | |
2534 | * we count 0 as well) we have to lower the s_maxbytes by one fs block. | |
2535 | * | |
cd2291a4 ES |
2536 | * Note, this does *not* consider any metadata overhead for vfs i_blocks. |
2537 | */ | |
f287a1a5 | 2538 | static loff_t ext4_max_size(int blkbits, int has_huge_files) |
cd2291a4 ES |
2539 | { |
2540 | loff_t res; | |
2541 | loff_t upper_limit = MAX_LFS_FILESIZE; | |
2542 | ||
2543 | /* small i_blocks in vfs inode? */ | |
f287a1a5 | 2544 | if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { |
cd2291a4 | 2545 | /* |
90c699a9 | 2546 | * CONFIG_LBDAF is not enabled implies the inode |
cd2291a4 ES |
2547 | * i_block represent total blocks in 512 bytes |
2548 | * 32 == size of vfs inode i_blocks * 8 | |
2549 | */ | |
2550 | upper_limit = (1LL << 32) - 1; | |
2551 | ||
2552 | /* total blocks in file system block size */ | |
2553 | upper_limit >>= (blkbits - 9); | |
2554 | upper_limit <<= blkbits; | |
2555 | } | |
2556 | ||
f17722f9 LC |
2557 | /* |
2558 | * 32-bit extent-start container, ee_block. We lower the maxbytes | |
2559 | * by one fs block, so ee_len can cover the extent of maximum file | |
2560 | * size | |
2561 | */ | |
2562 | res = (1LL << 32) - 1; | |
cd2291a4 | 2563 | res <<= blkbits; |
cd2291a4 ES |
2564 | |
2565 | /* Sanity check against vm- & vfs- imposed limits */ | |
2566 | if (res > upper_limit) | |
2567 | res = upper_limit; | |
2568 | ||
2569 | return res; | |
2570 | } | |
ac27a0ec | 2571 | |
ac27a0ec | 2572 | /* |
cd2291a4 | 2573 | * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect |
0fc1b451 AK |
2574 | * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. |
2575 | * We need to be 1 filesystem block less than the 2^48 sector limit. | |
ac27a0ec | 2576 | */ |
f287a1a5 | 2577 | static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) |
ac27a0ec | 2578 | { |
617ba13b | 2579 | loff_t res = EXT4_NDIR_BLOCKS; |
0fc1b451 AK |
2580 | int meta_blocks; |
2581 | loff_t upper_limit; | |
0b8e58a1 AD |
2582 | /* This is calculated to be the largest file size for a dense, block |
2583 | * mapped file such that the file's total number of 512-byte sectors, | |
2584 | * including data and all indirect blocks, does not exceed (2^48 - 1). | |
2585 | * | |
2586 | * __u32 i_blocks_lo and _u16 i_blocks_high represent the total | |
2587 | * number of 512-byte sectors of the file. | |
0fc1b451 AK |
2588 | */ |
2589 | ||
f287a1a5 | 2590 | if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { |
0fc1b451 | 2591 | /* |
90c699a9 | 2592 | * !has_huge_files or CONFIG_LBDAF not enabled implies that |
0b8e58a1 AD |
2593 | * the inode i_block field represents total file blocks in |
2594 | * 2^32 512-byte sectors == size of vfs inode i_blocks * 8 | |
0fc1b451 AK |
2595 | */ |
2596 | upper_limit = (1LL << 32) - 1; | |
2597 | ||
2598 | /* total blocks in file system block size */ | |
2599 | upper_limit >>= (bits - 9); | |
2600 | ||
2601 | } else { | |
8180a562 AK |
2602 | /* |
2603 | * We use 48 bit ext4_inode i_blocks | |
2604 | * With EXT4_HUGE_FILE_FL set the i_blocks | |
2605 | * represent total number of blocks in | |
2606 | * file system block size | |
2607 | */ | |
0fc1b451 AK |
2608 | upper_limit = (1LL << 48) - 1; |
2609 | ||
0fc1b451 AK |
2610 | } |
2611 | ||
2612 | /* indirect blocks */ | |
2613 | meta_blocks = 1; | |
2614 | /* double indirect blocks */ | |
2615 | meta_blocks += 1 + (1LL << (bits-2)); | |
2616 | /* tripple indirect blocks */ | |
2617 | meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); | |
2618 | ||
2619 | upper_limit -= meta_blocks; | |
2620 | upper_limit <<= bits; | |
ac27a0ec DK |
2621 | |
2622 | res += 1LL << (bits-2); | |
2623 | res += 1LL << (2*(bits-2)); | |
2624 | res += 1LL << (3*(bits-2)); | |
2625 | res <<= bits; | |
2626 | if (res > upper_limit) | |
2627 | res = upper_limit; | |
0fc1b451 AK |
2628 | |
2629 | if (res > MAX_LFS_FILESIZE) | |
2630 | res = MAX_LFS_FILESIZE; | |
2631 | ||
ac27a0ec DK |
2632 | return res; |
2633 | } | |
2634 | ||
617ba13b | 2635 | static ext4_fsblk_t descriptor_loc(struct super_block *sb, |
0b8e58a1 | 2636 | ext4_fsblk_t logical_sb_block, int nr) |
ac27a0ec | 2637 | { |
617ba13b | 2638 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
fd2d4291 | 2639 | ext4_group_t bg, first_meta_bg; |
ac27a0ec DK |
2640 | int has_super = 0; |
2641 | ||
2642 | first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); | |
2643 | ||
e2b911c5 | 2644 | if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) |
70bbb3e0 | 2645 | return logical_sb_block + nr + 1; |
ac27a0ec | 2646 | bg = sbi->s_desc_per_block * nr; |
617ba13b | 2647 | if (ext4_bg_has_super(sb, bg)) |
ac27a0ec | 2648 | has_super = 1; |
0b8e58a1 | 2649 | |
bd63f6b0 DW |
2650 | /* |
2651 | * If we have a meta_bg fs with 1k blocks, group 0's GDT is at | |
2652 | * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled | |
2653 | * on modern mke2fs or blksize > 1k on older mke2fs) then we must | |
2654 | * compensate. | |
2655 | */ | |
2656 | if (sb->s_blocksize == 1024 && nr == 0 && | |
2657 | le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0) | |
2658 | has_super++; | |
2659 | ||
617ba13b | 2660 | return (has_super + ext4_group_first_block_no(sb, bg)); |
ac27a0ec DK |
2661 | } |
2662 | ||
c9de560d AT |
2663 | /** |
2664 | * ext4_get_stripe_size: Get the stripe size. | |
2665 | * @sbi: In memory super block info | |
2666 | * | |
2667 | * If we have specified it via mount option, then | |
2668 | * use the mount option value. If the value specified at mount time is | |
2669 | * greater than the blocks per group use the super block value. | |
2670 | * If the super block value is greater than blocks per group return 0. | |
2671 | * Allocator needs it be less than blocks per group. | |
2672 | * | |
2673 | */ | |
2674 | static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) | |
2675 | { | |
2676 | unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); | |
2677 | unsigned long stripe_width = | |
2678 | le32_to_cpu(sbi->s_es->s_raid_stripe_width); | |
3eb08658 | 2679 | int ret; |
c9de560d AT |
2680 | |
2681 | if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) | |
3eb08658 | 2682 | ret = sbi->s_stripe; |
5469d7c3 | 2683 | else if (stripe_width && stripe_width <= sbi->s_blocks_per_group) |
3eb08658 | 2684 | ret = stripe_width; |
5469d7c3 | 2685 | else if (stride && stride <= sbi->s_blocks_per_group) |
3eb08658 DE |
2686 | ret = stride; |
2687 | else | |
2688 | ret = 0; | |
c9de560d | 2689 | |
3eb08658 DE |
2690 | /* |
2691 | * If the stripe width is 1, this makes no sense and | |
2692 | * we set it to 0 to turn off stripe handling code. | |
2693 | */ | |
2694 | if (ret <= 1) | |
2695 | ret = 0; | |
c9de560d | 2696 | |
3eb08658 | 2697 | return ret; |
c9de560d | 2698 | } |
ac27a0ec | 2699 | |
a13fb1a4 ES |
2700 | /* |
2701 | * Check whether this filesystem can be mounted based on | |
2702 | * the features present and the RDONLY/RDWR mount requested. | |
2703 | * Returns 1 if this filesystem can be mounted as requested, | |
2704 | * 0 if it cannot be. | |
2705 | */ | |
2706 | static int ext4_feature_set_ok(struct super_block *sb, int readonly) | |
2707 | { | |
e2b911c5 | 2708 | if (ext4_has_unknown_ext4_incompat_features(sb)) { |
a13fb1a4 ES |
2709 | ext4_msg(sb, KERN_ERR, |
2710 | "Couldn't mount because of " | |
2711 | "unsupported optional features (%x)", | |
2712 | (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & | |
2713 | ~EXT4_FEATURE_INCOMPAT_SUPP)); | |
2714 | return 0; | |
2715 | } | |
2716 | ||
2717 | if (readonly) | |
2718 | return 1; | |
2719 | ||
e2b911c5 | 2720 | if (ext4_has_feature_readonly(sb)) { |
2cb5cc8b DW |
2721 | ext4_msg(sb, KERN_INFO, "filesystem is read-only"); |
2722 | sb->s_flags |= MS_RDONLY; | |
2723 | return 1; | |
2724 | } | |
2725 | ||
a13fb1a4 | 2726 | /* Check that feature set is OK for a read-write mount */ |
e2b911c5 | 2727 | if (ext4_has_unknown_ext4_ro_compat_features(sb)) { |
a13fb1a4 ES |
2728 | ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " |
2729 | "unsupported optional features (%x)", | |
2730 | (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & | |
2731 | ~EXT4_FEATURE_RO_COMPAT_SUPP)); | |
2732 | return 0; | |
2733 | } | |
2734 | /* | |
2735 | * Large file size enabled file system can only be mounted | |
2736 | * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF | |
2737 | */ | |
e2b911c5 | 2738 | if (ext4_has_feature_huge_file(sb)) { |
a13fb1a4 ES |
2739 | if (sizeof(blkcnt_t) < sizeof(u64)) { |
2740 | ext4_msg(sb, KERN_ERR, "Filesystem with huge files " | |
2741 | "cannot be mounted RDWR without " | |
2742 | "CONFIG_LBDAF"); | |
2743 | return 0; | |
2744 | } | |
2745 | } | |
e2b911c5 | 2746 | if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { |
bab08ab9 TT |
2747 | ext4_msg(sb, KERN_ERR, |
2748 | "Can't support bigalloc feature without " | |
2749 | "extents feature\n"); | |
2750 | return 0; | |
2751 | } | |
7c319d32 AK |
2752 | |
2753 | #ifndef CONFIG_QUOTA | |
e2b911c5 | 2754 | if (ext4_has_feature_quota(sb) && !readonly) { |
7c319d32 AK |
2755 | ext4_msg(sb, KERN_ERR, |
2756 | "Filesystem with quota feature cannot be mounted RDWR " | |
2757 | "without CONFIG_QUOTA"); | |
2758 | return 0; | |
2759 | } | |
689c958c LX |
2760 | if (ext4_has_feature_project(sb) && !readonly) { |
2761 | ext4_msg(sb, KERN_ERR, | |
2762 | "Filesystem with project quota feature cannot be mounted RDWR " | |
2763 | "without CONFIG_QUOTA"); | |
2764 | return 0; | |
2765 | } | |
7c319d32 | 2766 | #endif /* CONFIG_QUOTA */ |
a13fb1a4 ES |
2767 | return 1; |
2768 | } | |
2769 | ||
66e61a9e TT |
2770 | /* |
2771 | * This function is called once a day if we have errors logged | |
2772 | * on the file system | |
2773 | */ | |
2774 | static void print_daily_error_info(unsigned long arg) | |
2775 | { | |
2776 | struct super_block *sb = (struct super_block *) arg; | |
2777 | struct ext4_sb_info *sbi; | |
2778 | struct ext4_super_block *es; | |
2779 | ||
2780 | sbi = EXT4_SB(sb); | |
2781 | es = sbi->s_es; | |
2782 | ||
2783 | if (es->s_error_count) | |
ae0f78de TT |
2784 | /* fsck newer than v1.41.13 is needed to clean this condition. */ |
2785 | ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", | |
66e61a9e TT |
2786 | le32_to_cpu(es->s_error_count)); |
2787 | if (es->s_first_error_time) { | |
ae0f78de | 2788 | printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d", |
66e61a9e TT |
2789 | sb->s_id, le32_to_cpu(es->s_first_error_time), |
2790 | (int) sizeof(es->s_first_error_func), | |
2791 | es->s_first_error_func, | |
2792 | le32_to_cpu(es->s_first_error_line)); | |
2793 | if (es->s_first_error_ino) | |
651e1c3b | 2794 | printk(KERN_CONT ": inode %u", |
66e61a9e TT |
2795 | le32_to_cpu(es->s_first_error_ino)); |
2796 | if (es->s_first_error_block) | |
651e1c3b | 2797 | printk(KERN_CONT ": block %llu", (unsigned long long) |
66e61a9e | 2798 | le64_to_cpu(es->s_first_error_block)); |
651e1c3b | 2799 | printk(KERN_CONT "\n"); |
66e61a9e TT |
2800 | } |
2801 | if (es->s_last_error_time) { | |
ae0f78de | 2802 | printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d", |
66e61a9e TT |
2803 | sb->s_id, le32_to_cpu(es->s_last_error_time), |
2804 | (int) sizeof(es->s_last_error_func), | |
2805 | es->s_last_error_func, | |
2806 | le32_to_cpu(es->s_last_error_line)); | |
2807 | if (es->s_last_error_ino) | |
651e1c3b | 2808 | printk(KERN_CONT ": inode %u", |
66e61a9e TT |
2809 | le32_to_cpu(es->s_last_error_ino)); |
2810 | if (es->s_last_error_block) | |
651e1c3b | 2811 | printk(KERN_CONT ": block %llu", (unsigned long long) |
66e61a9e | 2812 | le64_to_cpu(es->s_last_error_block)); |
651e1c3b | 2813 | printk(KERN_CONT "\n"); |
66e61a9e TT |
2814 | } |
2815 | mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ | |
2816 | } | |
2817 | ||
bfff6873 LC |
2818 | /* Find next suitable group and run ext4_init_inode_table */ |
2819 | static int ext4_run_li_request(struct ext4_li_request *elr) | |
2820 | { | |
2821 | struct ext4_group_desc *gdp = NULL; | |
2822 | ext4_group_t group, ngroups; | |
2823 | struct super_block *sb; | |
2824 | unsigned long timeout = 0; | |
2825 | int ret = 0; | |
2826 | ||
2827 | sb = elr->lr_super; | |
2828 | ngroups = EXT4_SB(sb)->s_groups_count; | |
2829 | ||
2830 | for (group = elr->lr_next_group; group < ngroups; group++) { | |
2831 | gdp = ext4_get_group_desc(sb, group, NULL); | |
2832 | if (!gdp) { | |
2833 | ret = 1; | |
2834 | break; | |
2835 | } | |
2836 | ||
2837 | if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) | |
2838 | break; | |
2839 | } | |
2840 | ||
7f511862 | 2841 | if (group >= ngroups) |
bfff6873 LC |
2842 | ret = 1; |
2843 | ||
2844 | if (!ret) { | |
2845 | timeout = jiffies; | |
2846 | ret = ext4_init_inode_table(sb, group, | |
2847 | elr->lr_timeout ? 0 : 1); | |
2848 | if (elr->lr_timeout == 0) { | |
51ce6511 LC |
2849 | timeout = (jiffies - timeout) * |
2850 | elr->lr_sbi->s_li_wait_mult; | |
bfff6873 LC |
2851 | elr->lr_timeout = timeout; |
2852 | } | |
2853 | elr->lr_next_sched = jiffies + elr->lr_timeout; | |
2854 | elr->lr_next_group = group + 1; | |
2855 | } | |
bfff6873 LC |
2856 | return ret; |
2857 | } | |
2858 | ||
2859 | /* | |
2860 | * Remove lr_request from the list_request and free the | |
4ed5c033 | 2861 | * request structure. Should be called with li_list_mtx held |
bfff6873 LC |
2862 | */ |
2863 | static void ext4_remove_li_request(struct ext4_li_request *elr) | |
2864 | { | |
2865 | struct ext4_sb_info *sbi; | |
2866 | ||
2867 | if (!elr) | |
2868 | return; | |
2869 | ||
2870 | sbi = elr->lr_sbi; | |
2871 | ||
2872 | list_del(&elr->lr_request); | |
2873 | sbi->s_li_request = NULL; | |
2874 | kfree(elr); | |
2875 | } | |
2876 | ||
2877 | static void ext4_unregister_li_request(struct super_block *sb) | |
2878 | { | |
1bb933fb LC |
2879 | mutex_lock(&ext4_li_mtx); |
2880 | if (!ext4_li_info) { | |
2881 | mutex_unlock(&ext4_li_mtx); | |
bfff6873 | 2882 | return; |
1bb933fb | 2883 | } |
bfff6873 LC |
2884 | |
2885 | mutex_lock(&ext4_li_info->li_list_mtx); | |
1bb933fb | 2886 | ext4_remove_li_request(EXT4_SB(sb)->s_li_request); |
bfff6873 | 2887 | mutex_unlock(&ext4_li_info->li_list_mtx); |
1bb933fb | 2888 | mutex_unlock(&ext4_li_mtx); |
bfff6873 LC |
2889 | } |
2890 | ||
8f1f7453 ES |
2891 | static struct task_struct *ext4_lazyinit_task; |
2892 | ||
bfff6873 LC |
2893 | /* |
2894 | * This is the function where ext4lazyinit thread lives. It walks | |
2895 | * through the request list searching for next scheduled filesystem. | |
2896 | * When such a fs is found, run the lazy initialization request | |
2897 | * (ext4_rn_li_request) and keep track of the time spend in this | |
2898 | * function. Based on that time we compute next schedule time of | |
2899 | * the request. When walking through the list is complete, compute | |
2900 | * next waking time and put itself into sleep. | |
2901 | */ | |
2902 | static int ext4_lazyinit_thread(void *arg) | |
2903 | { | |
2904 | struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; | |
2905 | struct list_head *pos, *n; | |
2906 | struct ext4_li_request *elr; | |
4ed5c033 | 2907 | unsigned long next_wakeup, cur; |
bfff6873 LC |
2908 | |
2909 | BUG_ON(NULL == eli); | |
2910 | ||
bfff6873 LC |
2911 | cont_thread: |
2912 | while (true) { | |
2913 | next_wakeup = MAX_JIFFY_OFFSET; | |
2914 | ||
2915 | mutex_lock(&eli->li_list_mtx); | |
2916 | if (list_empty(&eli->li_request_list)) { | |
2917 | mutex_unlock(&eli->li_list_mtx); | |
2918 | goto exit_thread; | |
2919 | } | |
bfff6873 | 2920 | list_for_each_safe(pos, n, &eli->li_request_list) { |
e22834f0 DM |
2921 | int err = 0; |
2922 | int progress = 0; | |
bfff6873 LC |
2923 | elr = list_entry(pos, struct ext4_li_request, |
2924 | lr_request); | |
2925 | ||
e22834f0 DM |
2926 | if (time_before(jiffies, elr->lr_next_sched)) { |
2927 | if (time_before(elr->lr_next_sched, next_wakeup)) | |
2928 | next_wakeup = elr->lr_next_sched; | |
2929 | continue; | |
2930 | } | |
2931 | if (down_read_trylock(&elr->lr_super->s_umount)) { | |
2932 | if (sb_start_write_trylock(elr->lr_super)) { | |
2933 | progress = 1; | |
2934 | /* | |
2935 | * We hold sb->s_umount, sb can not | |
2936 | * be removed from the list, it is | |
2937 | * now safe to drop li_list_mtx | |
2938 | */ | |
2939 | mutex_unlock(&eli->li_list_mtx); | |
2940 | err = ext4_run_li_request(elr); | |
2941 | sb_end_write(elr->lr_super); | |
2942 | mutex_lock(&eli->li_list_mtx); | |
2943 | n = pos->next; | |
b2c78cd0 | 2944 | } |
e22834f0 DM |
2945 | up_read((&elr->lr_super->s_umount)); |
2946 | } | |
2947 | /* error, remove the lazy_init job */ | |
2948 | if (err) { | |
2949 | ext4_remove_li_request(elr); | |
2950 | continue; | |
2951 | } | |
2952 | if (!progress) { | |
2953 | elr->lr_next_sched = jiffies + | |
2954 | (prandom_u32() | |
2955 | % (EXT4_DEF_LI_MAX_START_DELAY * HZ)); | |
bfff6873 | 2956 | } |
bfff6873 LC |
2957 | if (time_before(elr->lr_next_sched, next_wakeup)) |
2958 | next_wakeup = elr->lr_next_sched; | |
2959 | } | |
2960 | mutex_unlock(&eli->li_list_mtx); | |
2961 | ||
a0acae0e | 2962 | try_to_freeze(); |
bfff6873 | 2963 | |
4ed5c033 LC |
2964 | cur = jiffies; |
2965 | if ((time_after_eq(cur, next_wakeup)) || | |
f4245bd4 | 2966 | (MAX_JIFFY_OFFSET == next_wakeup)) { |
bfff6873 LC |
2967 | cond_resched(); |
2968 | continue; | |
2969 | } | |
2970 | ||
4ed5c033 LC |
2971 | schedule_timeout_interruptible(next_wakeup - cur); |
2972 | ||
8f1f7453 ES |
2973 | if (kthread_should_stop()) { |
2974 | ext4_clear_request_list(); | |
2975 | goto exit_thread; | |
2976 | } | |
bfff6873 LC |
2977 | } |
2978 | ||
2979 | exit_thread: | |
2980 | /* | |
2981 | * It looks like the request list is empty, but we need | |
2982 | * to check it under the li_list_mtx lock, to prevent any | |
2983 | * additions into it, and of course we should lock ext4_li_mtx | |
2984 | * to atomically free the list and ext4_li_info, because at | |
2985 | * this point another ext4 filesystem could be registering | |
2986 | * new one. | |
2987 | */ | |
2988 | mutex_lock(&ext4_li_mtx); | |
2989 | mutex_lock(&eli->li_list_mtx); | |
2990 | if (!list_empty(&eli->li_request_list)) { | |
2991 | mutex_unlock(&eli->li_list_mtx); | |
2992 | mutex_unlock(&ext4_li_mtx); | |
2993 | goto cont_thread; | |
2994 | } | |
2995 | mutex_unlock(&eli->li_list_mtx); | |
bfff6873 LC |
2996 | kfree(ext4_li_info); |
2997 | ext4_li_info = NULL; | |
2998 | mutex_unlock(&ext4_li_mtx); | |
2999 | ||
3000 | return 0; | |
3001 | } | |
3002 | ||
3003 | static void ext4_clear_request_list(void) | |
3004 | { | |
3005 | struct list_head *pos, *n; | |
3006 | struct ext4_li_request *elr; | |
3007 | ||
3008 | mutex_lock(&ext4_li_info->li_list_mtx); | |
bfff6873 LC |
3009 | list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { |
3010 | elr = list_entry(pos, struct ext4_li_request, | |
3011 | lr_request); | |
3012 | ext4_remove_li_request(elr); | |
3013 | } | |
3014 | mutex_unlock(&ext4_li_info->li_list_mtx); | |
3015 | } | |
3016 | ||
3017 | static int ext4_run_lazyinit_thread(void) | |
3018 | { | |
8f1f7453 ES |
3019 | ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, |
3020 | ext4_li_info, "ext4lazyinit"); | |
3021 | if (IS_ERR(ext4_lazyinit_task)) { | |
3022 | int err = PTR_ERR(ext4_lazyinit_task); | |
bfff6873 | 3023 | ext4_clear_request_list(); |
bfff6873 LC |
3024 | kfree(ext4_li_info); |
3025 | ext4_li_info = NULL; | |
92b97816 | 3026 | printk(KERN_CRIT "EXT4-fs: error %d creating inode table " |
bfff6873 LC |
3027 | "initialization thread\n", |
3028 | err); | |
3029 | return err; | |
3030 | } | |
3031 | ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; | |
bfff6873 LC |
3032 | return 0; |
3033 | } | |
3034 | ||
3035 | /* | |
3036 | * Check whether it make sense to run itable init. thread or not. | |
3037 | * If there is at least one uninitialized inode table, return | |
3038 | * corresponding group number, else the loop goes through all | |
3039 | * groups and return total number of groups. | |
3040 | */ | |
3041 | static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) | |
3042 | { | |
3043 | ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; | |
3044 | struct ext4_group_desc *gdp = NULL; | |
3045 | ||
3046 | for (group = 0; group < ngroups; group++) { | |
3047 | gdp = ext4_get_group_desc(sb, group, NULL); | |
3048 | if (!gdp) | |
3049 | continue; | |
3050 | ||
3051 | if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) | |
3052 | break; | |
3053 | } | |
3054 | ||
3055 | return group; | |
3056 | } | |
3057 | ||
3058 | static int ext4_li_info_new(void) | |
3059 | { | |
3060 | struct ext4_lazy_init *eli = NULL; | |
3061 | ||
3062 | eli = kzalloc(sizeof(*eli), GFP_KERNEL); | |
3063 | if (!eli) | |
3064 | return -ENOMEM; | |
3065 | ||
bfff6873 LC |
3066 | INIT_LIST_HEAD(&eli->li_request_list); |
3067 | mutex_init(&eli->li_list_mtx); | |
3068 | ||
bfff6873 LC |
3069 | eli->li_state |= EXT4_LAZYINIT_QUIT; |
3070 | ||
3071 | ext4_li_info = eli; | |
3072 | ||
3073 | return 0; | |
3074 | } | |
3075 | ||
3076 | static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, | |
3077 | ext4_group_t start) | |
3078 | { | |
3079 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
3080 | struct ext4_li_request *elr; | |
bfff6873 LC |
3081 | |
3082 | elr = kzalloc(sizeof(*elr), GFP_KERNEL); | |
3083 | if (!elr) | |
3084 | return NULL; | |
3085 | ||
3086 | elr->lr_super = sb; | |
3087 | elr->lr_sbi = sbi; | |
3088 | elr->lr_next_group = start; | |
3089 | ||
3090 | /* | |
3091 | * Randomize first schedule time of the request to | |
3092 | * spread the inode table initialization requests | |
3093 | * better. | |
3094 | */ | |
dd1f723b TT |
3095 | elr->lr_next_sched = jiffies + (prandom_u32() % |
3096 | (EXT4_DEF_LI_MAX_START_DELAY * HZ)); | |
bfff6873 LC |
3097 | return elr; |
3098 | } | |
3099 | ||
7f511862 TT |
3100 | int ext4_register_li_request(struct super_block *sb, |
3101 | ext4_group_t first_not_zeroed) | |
bfff6873 LC |
3102 | { |
3103 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
7f511862 | 3104 | struct ext4_li_request *elr = NULL; |
bfff6873 | 3105 | ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; |
6c5a6cb9 | 3106 | int ret = 0; |
bfff6873 | 3107 | |
7f511862 | 3108 | mutex_lock(&ext4_li_mtx); |
51ce6511 LC |
3109 | if (sbi->s_li_request != NULL) { |
3110 | /* | |
3111 | * Reset timeout so it can be computed again, because | |
3112 | * s_li_wait_mult might have changed. | |
3113 | */ | |
3114 | sbi->s_li_request->lr_timeout = 0; | |
7f511862 | 3115 | goto out; |
51ce6511 | 3116 | } |
bfff6873 | 3117 | |
bc98a42c | 3118 | if (first_not_zeroed == ngroups || sb_rdonly(sb) || |
55ff3840 | 3119 | !test_opt(sb, INIT_INODE_TABLE)) |
7f511862 | 3120 | goto out; |
bfff6873 LC |
3121 | |
3122 | elr = ext4_li_request_new(sb, first_not_zeroed); | |
7f511862 TT |
3123 | if (!elr) { |
3124 | ret = -ENOMEM; | |
3125 | goto out; | |
3126 | } | |
bfff6873 LC |
3127 | |
3128 | if (NULL == ext4_li_info) { | |
3129 | ret = ext4_li_info_new(); | |
3130 | if (ret) | |
3131 | goto out; | |
3132 | } | |
3133 | ||
3134 | mutex_lock(&ext4_li_info->li_list_mtx); | |
3135 | list_add(&elr->lr_request, &ext4_li_info->li_request_list); | |
3136 | mutex_unlock(&ext4_li_info->li_list_mtx); | |
3137 | ||
3138 | sbi->s_li_request = elr; | |
46e4690b TM |
3139 | /* |
3140 | * set elr to NULL here since it has been inserted to | |
3141 | * the request_list and the removal and free of it is | |
3142 | * handled by ext4_clear_request_list from now on. | |
3143 | */ | |
3144 | elr = NULL; | |
bfff6873 LC |
3145 | |
3146 | if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { | |
3147 | ret = ext4_run_lazyinit_thread(); | |
3148 | if (ret) | |
3149 | goto out; | |
3150 | } | |
bfff6873 | 3151 | out: |
beed5ecb NK |
3152 | mutex_unlock(&ext4_li_mtx); |
3153 | if (ret) | |
bfff6873 | 3154 | kfree(elr); |
bfff6873 LC |
3155 | return ret; |
3156 | } | |
3157 | ||
3158 | /* | |
3159 | * We do not need to lock anything since this is called on | |
3160 | * module unload. | |
3161 | */ | |
3162 | static void ext4_destroy_lazyinit_thread(void) | |
3163 | { | |
3164 | /* | |
3165 | * If thread exited earlier | |
3166 | * there's nothing to be done. | |
3167 | */ | |
8f1f7453 | 3168 | if (!ext4_li_info || !ext4_lazyinit_task) |
bfff6873 LC |
3169 | return; |
3170 | ||
8f1f7453 | 3171 | kthread_stop(ext4_lazyinit_task); |
bfff6873 LC |
3172 | } |
3173 | ||
25ed6e8a DW |
3174 | static int set_journal_csum_feature_set(struct super_block *sb) |
3175 | { | |
3176 | int ret = 1; | |
3177 | int compat, incompat; | |
3178 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
3179 | ||
9aa5d32b | 3180 | if (ext4_has_metadata_csum(sb)) { |
db9ee220 | 3181 | /* journal checksum v3 */ |
25ed6e8a | 3182 | compat = 0; |
db9ee220 | 3183 | incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; |
25ed6e8a DW |
3184 | } else { |
3185 | /* journal checksum v1 */ | |
3186 | compat = JBD2_FEATURE_COMPAT_CHECKSUM; | |
3187 | incompat = 0; | |
3188 | } | |
3189 | ||
feb8c6d3 DW |
3190 | jbd2_journal_clear_features(sbi->s_journal, |
3191 | JBD2_FEATURE_COMPAT_CHECKSUM, 0, | |
3192 | JBD2_FEATURE_INCOMPAT_CSUM_V3 | | |
3193 | JBD2_FEATURE_INCOMPAT_CSUM_V2); | |
25ed6e8a DW |
3194 | if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { |
3195 | ret = jbd2_journal_set_features(sbi->s_journal, | |
3196 | compat, 0, | |
3197 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | | |
3198 | incompat); | |
3199 | } else if (test_opt(sb, JOURNAL_CHECKSUM)) { | |
3200 | ret = jbd2_journal_set_features(sbi->s_journal, | |
3201 | compat, 0, | |
3202 | incompat); | |
3203 | jbd2_journal_clear_features(sbi->s_journal, 0, 0, | |
3204 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); | |
3205 | } else { | |
feb8c6d3 DW |
3206 | jbd2_journal_clear_features(sbi->s_journal, 0, 0, |
3207 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); | |
25ed6e8a DW |
3208 | } |
3209 | ||
3210 | return ret; | |
3211 | } | |
3212 | ||
952fc18e TT |
3213 | /* |
3214 | * Note: calculating the overhead so we can be compatible with | |
3215 | * historical BSD practice is quite difficult in the face of | |
3216 | * clusters/bigalloc. This is because multiple metadata blocks from | |
3217 | * different block group can end up in the same allocation cluster. | |
3218 | * Calculating the exact overhead in the face of clustered allocation | |
3219 | * requires either O(all block bitmaps) in memory or O(number of block | |
3220 | * groups**2) in time. We will still calculate the superblock for | |
3221 | * older file systems --- and if we come across with a bigalloc file | |
3222 | * system with zero in s_overhead_clusters the estimate will be close to | |
3223 | * correct especially for very large cluster sizes --- but for newer | |
3224 | * file systems, it's better to calculate this figure once at mkfs | |
3225 | * time, and store it in the superblock. If the superblock value is | |
3226 | * present (even for non-bigalloc file systems), we will use it. | |
3227 | */ | |
3228 | static int count_overhead(struct super_block *sb, ext4_group_t grp, | |
3229 | char *buf) | |
3230 | { | |
3231 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
3232 | struct ext4_group_desc *gdp; | |
3233 | ext4_fsblk_t first_block, last_block, b; | |
3234 | ext4_group_t i, ngroups = ext4_get_groups_count(sb); | |
3235 | int s, j, count = 0; | |
3236 | ||
e2b911c5 | 3237 | if (!ext4_has_feature_bigalloc(sb)) |
0548bbb8 TT |
3238 | return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) + |
3239 | sbi->s_itb_per_group + 2); | |
3240 | ||
952fc18e TT |
3241 | first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + |
3242 | (grp * EXT4_BLOCKS_PER_GROUP(sb)); | |
3243 | last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; | |
3244 | for (i = 0; i < ngroups; i++) { | |
3245 | gdp = ext4_get_group_desc(sb, i, NULL); | |
3246 | b = ext4_block_bitmap(sb, gdp); | |
3247 | if (b >= first_block && b <= last_block) { | |
3248 | ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); | |
3249 | count++; | |
3250 | } | |
3251 | b = ext4_inode_bitmap(sb, gdp); | |
3252 | if (b >= first_block && b <= last_block) { | |
3253 | ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); | |
3254 | count++; | |
3255 | } | |
3256 | b = ext4_inode_table(sb, gdp); | |
3257 | if (b >= first_block && b + sbi->s_itb_per_group <= last_block) | |
3258 | for (j = 0; j < sbi->s_itb_per_group; j++, b++) { | |
3259 | int c = EXT4_B2C(sbi, b - first_block); | |
3260 | ext4_set_bit(c, buf); | |
3261 | count++; | |
3262 | } | |
3263 | if (i != grp) | |
3264 | continue; | |
3265 | s = 0; | |
3266 | if (ext4_bg_has_super(sb, grp)) { | |
3267 | ext4_set_bit(s++, buf); | |
3268 | count++; | |
3269 | } | |
c48ae41b TT |
3270 | j = ext4_bg_num_gdb(sb, grp); |
3271 | if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) { | |
3272 | ext4_error(sb, "Invalid number of block group " | |
3273 | "descriptor blocks: %d", j); | |
3274 | j = EXT4_BLOCKS_PER_GROUP(sb) - s; | |
952fc18e | 3275 | } |
c48ae41b TT |
3276 | count += j; |
3277 | for (; j > 0; j--) | |
3278 | ext4_set_bit(EXT4_B2C(sbi, s++), buf); | |
952fc18e TT |
3279 | } |
3280 | if (!count) | |
3281 | return 0; | |
3282 | return EXT4_CLUSTERS_PER_GROUP(sb) - | |
3283 | ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); | |
3284 | } | |
3285 | ||
3286 | /* | |
3287 | * Compute the overhead and stash it in sbi->s_overhead | |
3288 | */ | |
3289 | int ext4_calculate_overhead(struct super_block *sb) | |
3290 | { | |
3291 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
3292 | struct ext4_super_block *es = sbi->s_es; | |
3c816ded EW |
3293 | struct inode *j_inode; |
3294 | unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum); | |
952fc18e TT |
3295 | ext4_group_t i, ngroups = ext4_get_groups_count(sb); |
3296 | ext4_fsblk_t overhead = 0; | |
4fdb5543 | 3297 | char *buf = (char *) get_zeroed_page(GFP_NOFS); |
952fc18e | 3298 | |
952fc18e TT |
3299 | if (!buf) |
3300 | return -ENOMEM; | |
3301 | ||
3302 | /* | |
3303 | * Compute the overhead (FS structures). This is constant | |
3304 | * for a given filesystem unless the number of block groups | |
3305 | * changes so we cache the previous value until it does. | |
3306 | */ | |
3307 | ||
3308 | /* | |
3309 | * All of the blocks before first_data_block are overhead | |
3310 | */ | |
3311 | overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); | |
3312 | ||
3313 | /* | |
3314 | * Add the overhead found in each block group | |
3315 | */ | |
3316 | for (i = 0; i < ngroups; i++) { | |
3317 | int blks; | |
3318 | ||
3319 | blks = count_overhead(sb, i, buf); | |
3320 | overhead += blks; | |
3321 | if (blks) | |
3322 | memset(buf, 0, PAGE_SIZE); | |
3323 | cond_resched(); | |
3324 | } | |
3c816ded EW |
3325 | |
3326 | /* | |
3327 | * Add the internal journal blocks whether the journal has been | |
3328 | * loaded or not | |
3329 | */ | |
b003b524 | 3330 | if (sbi->s_journal && !sbi->journal_bdev) |
810da240 | 3331 | overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen); |
3c816ded EW |
3332 | else if (ext4_has_feature_journal(sb) && !sbi->s_journal) { |
3333 | j_inode = ext4_get_journal_inode(sb, j_inum); | |
3334 | if (j_inode) { | |
3335 | j_blocks = j_inode->i_size >> sb->s_blocksize_bits; | |
3336 | overhead += EXT4_NUM_B2C(sbi, j_blocks); | |
3337 | iput(j_inode); | |
3338 | } else { | |
3339 | ext4_msg(sb, KERN_ERR, "can't get journal size"); | |
3340 | } | |
3341 | } | |
952fc18e TT |
3342 | sbi->s_overhead = overhead; |
3343 | smp_wmb(); | |
3344 | free_page((unsigned long) buf); | |
3345 | return 0; | |
3346 | } | |
3347 | ||
b5799018 | 3348 | static void ext4_set_resv_clusters(struct super_block *sb) |
27dd4385 LC |
3349 | { |
3350 | ext4_fsblk_t resv_clusters; | |
b5799018 | 3351 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
27dd4385 | 3352 | |
30fac0f7 JK |
3353 | /* |
3354 | * There's no need to reserve anything when we aren't using extents. | |
3355 | * The space estimates are exact, there are no unwritten extents, | |
3356 | * hole punching doesn't need new metadata... This is needed especially | |
3357 | * to keep ext2/3 backward compatibility. | |
3358 | */ | |
e2b911c5 | 3359 | if (!ext4_has_feature_extents(sb)) |
b5799018 | 3360 | return; |
27dd4385 LC |
3361 | /* |
3362 | * By default we reserve 2% or 4096 clusters, whichever is smaller. | |
3363 | * This should cover the situations where we can not afford to run | |
3364 | * out of space like for example punch hole, or converting | |
556615dc | 3365 | * unwritten extents in delalloc path. In most cases such |
27dd4385 LC |
3366 | * allocation would require 1, or 2 blocks, higher numbers are |
3367 | * very rare. | |
3368 | */ | |
b5799018 TT |
3369 | resv_clusters = (ext4_blocks_count(sbi->s_es) >> |
3370 | sbi->s_cluster_bits); | |
27dd4385 LC |
3371 | |
3372 | do_div(resv_clusters, 50); | |
3373 | resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); | |
3374 | ||
b5799018 | 3375 | atomic64_set(&sbi->s_resv_clusters, resv_clusters); |
27dd4385 LC |
3376 | } |
3377 | ||
2b2d6d01 | 3378 | static int ext4_fill_super(struct super_block *sb, void *data, int silent) |
ac27a0ec | 3379 | { |
d4c402d9 | 3380 | char *orig_data = kstrdup(data, GFP_KERNEL); |
2b2d6d01 | 3381 | struct buffer_head *bh; |
617ba13b | 3382 | struct ext4_super_block *es = NULL; |
5aee0f8a | 3383 | struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); |
617ba13b MC |
3384 | ext4_fsblk_t block; |
3385 | ext4_fsblk_t sb_block = get_sb_block(&data); | |
70bbb3e0 | 3386 | ext4_fsblk_t logical_sb_block; |
ac27a0ec | 3387 | unsigned long offset = 0; |
ac27a0ec DK |
3388 | unsigned long journal_devnum = 0; |
3389 | unsigned long def_mount_opts; | |
3390 | struct inode *root; | |
0390131b | 3391 | const char *descr; |
dcc7dae3 | 3392 | int ret = -ENOMEM; |
281b5995 | 3393 | int blocksize, clustersize; |
4ec11028 TT |
3394 | unsigned int db_count; |
3395 | unsigned int i; | |
281b5995 | 3396 | int needs_recovery, has_huge_files, has_bigalloc; |
bd81d8ee | 3397 | __u64 blocks_count; |
07aa2ea1 | 3398 | int err = 0; |
b3881f74 | 3399 | unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; |
bfff6873 | 3400 | ext4_group_t first_not_zeroed; |
ac27a0ec | 3401 | |
5aee0f8a TT |
3402 | if ((data && !orig_data) || !sbi) |
3403 | goto out_free_base; | |
705895b6 PE |
3404 | |
3405 | sbi->s_blockgroup_lock = | |
3406 | kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); | |
5aee0f8a TT |
3407 | if (!sbi->s_blockgroup_lock) |
3408 | goto out_free_base; | |
3409 | ||
ac27a0ec | 3410 | sb->s_fs_info = sbi; |
2c0544b2 | 3411 | sbi->s_sb = sb; |
240799cd | 3412 | sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; |
d9c9bef1 | 3413 | sbi->s_sb_block = sb_block; |
f613dfcb TT |
3414 | if (sb->s_bdev->bd_part) |
3415 | sbi->s_sectors_written_start = | |
3416 | part_stat_read(sb->s_bdev->bd_part, sectors[1]); | |
ac27a0ec | 3417 | |
9f6200bb | 3418 | /* Cleanup superblock name */ |
ec3904dc | 3419 | strreplace(sb->s_id, '/', '!'); |
9f6200bb | 3420 | |
07aa2ea1 | 3421 | /* -EINVAL is default */ |
dcc7dae3 | 3422 | ret = -EINVAL; |
617ba13b | 3423 | blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); |
ac27a0ec | 3424 | if (!blocksize) { |
b31e1552 | 3425 | ext4_msg(sb, KERN_ERR, "unable to set blocksize"); |
ac27a0ec DK |
3426 | goto out_fail; |
3427 | } | |
3428 | ||
3429 | /* | |
617ba13b | 3430 | * The ext4 superblock will not be buffer aligned for other than 1kB |
ac27a0ec DK |
3431 | * block sizes. We need to calculate the offset from buffer start. |
3432 | */ | |
617ba13b | 3433 | if (blocksize != EXT4_MIN_BLOCK_SIZE) { |
70bbb3e0 AM |
3434 | logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; |
3435 | offset = do_div(logical_sb_block, blocksize); | |
ac27a0ec | 3436 | } else { |
70bbb3e0 | 3437 | logical_sb_block = sb_block; |
ac27a0ec DK |
3438 | } |
3439 | ||
a8ac900b | 3440 | if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) { |
b31e1552 | 3441 | ext4_msg(sb, KERN_ERR, "unable to read superblock"); |
ac27a0ec DK |
3442 | goto out_fail; |
3443 | } | |
3444 | /* | |
3445 | * Note: s_es must be initialized as soon as possible because | |
617ba13b | 3446 | * some ext4 macro-instructions depend on its value |
ac27a0ec | 3447 | */ |
2716b802 | 3448 | es = (struct ext4_super_block *) (bh->b_data + offset); |
ac27a0ec DK |
3449 | sbi->s_es = es; |
3450 | sb->s_magic = le16_to_cpu(es->s_magic); | |
617ba13b MC |
3451 | if (sb->s_magic != EXT4_SUPER_MAGIC) |
3452 | goto cantfind_ext4; | |
afc32f7e | 3453 | sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); |
ac27a0ec | 3454 | |
feb0ab32 | 3455 | /* Warn if metadata_csum and gdt_csum are both set. */ |
e2b911c5 DW |
3456 | if (ext4_has_feature_metadata_csum(sb) && |
3457 | ext4_has_feature_gdt_csum(sb)) | |
363307e6 | 3458 | ext4_warning(sb, "metadata_csum and uninit_bg are " |
feb0ab32 DW |
3459 | "redundant flags; please run fsck."); |
3460 | ||
d25425f8 DW |
3461 | /* Check for a known checksum algorithm */ |
3462 | if (!ext4_verify_csum_type(sb, es)) { | |
3463 | ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " | |
3464 | "unknown checksum algorithm."); | |
3465 | silent = 1; | |
3466 | goto cantfind_ext4; | |
3467 | } | |
3468 | ||
0441984a | 3469 | /* Load the checksum driver */ |
dec214d0 TE |
3470 | if (ext4_has_feature_metadata_csum(sb) || |
3471 | ext4_has_feature_ea_inode(sb)) { | |
0441984a DW |
3472 | sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); |
3473 | if (IS_ERR(sbi->s_chksum_driver)) { | |
3474 | ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); | |
3475 | ret = PTR_ERR(sbi->s_chksum_driver); | |
3476 | sbi->s_chksum_driver = NULL; | |
3477 | goto failed_mount; | |
3478 | } | |
3479 | } | |
3480 | ||
a9c47317 DW |
3481 | /* Check superblock checksum */ |
3482 | if (!ext4_superblock_csum_verify(sb, es)) { | |
3483 | ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " | |
3484 | "invalid superblock checksum. Run e2fsck?"); | |
3485 | silent = 1; | |
6a797d27 | 3486 | ret = -EFSBADCRC; |
a9c47317 DW |
3487 | goto cantfind_ext4; |
3488 | } | |
3489 | ||
3490 | /* Precompute checksum seed for all metadata */ | |
e2b911c5 | 3491 | if (ext4_has_feature_csum_seed(sb)) |
8c81bd8f | 3492 | sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); |
dec214d0 | 3493 | else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb)) |
a9c47317 DW |
3494 | sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, |
3495 | sizeof(es->s_uuid)); | |
3496 | ||
ac27a0ec DK |
3497 | /* Set defaults before we parse the mount options */ |
3498 | def_mount_opts = le32_to_cpu(es->s_default_mount_opts); | |
fd8c37ec | 3499 | set_opt(sb, INIT_INODE_TABLE); |
617ba13b | 3500 | if (def_mount_opts & EXT4_DEFM_DEBUG) |
fd8c37ec | 3501 | set_opt(sb, DEBUG); |
87f26807 | 3502 | if (def_mount_opts & EXT4_DEFM_BSDGROUPS) |
fd8c37ec | 3503 | set_opt(sb, GRPID); |
617ba13b | 3504 | if (def_mount_opts & EXT4_DEFM_UID16) |
fd8c37ec | 3505 | set_opt(sb, NO_UID32); |
ea663336 | 3506 | /* xattr user namespace & acls are now defaulted on */ |
ea663336 | 3507 | set_opt(sb, XATTR_USER); |
03010a33 | 3508 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
ea663336 | 3509 | set_opt(sb, POSIX_ACL); |
2e7842b8 | 3510 | #endif |
98c1a759 DW |
3511 | /* don't forget to enable journal_csum when metadata_csum is enabled. */ |
3512 | if (ext4_has_metadata_csum(sb)) | |
3513 | set_opt(sb, JOURNAL_CHECKSUM); | |
3514 | ||
617ba13b | 3515 | if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) |
fd8c37ec | 3516 | set_opt(sb, JOURNAL_DATA); |
617ba13b | 3517 | else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) |
fd8c37ec | 3518 | set_opt(sb, ORDERED_DATA); |
617ba13b | 3519 | else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) |
fd8c37ec | 3520 | set_opt(sb, WRITEBACK_DATA); |
617ba13b MC |
3521 | |
3522 | if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) | |
fd8c37ec | 3523 | set_opt(sb, ERRORS_PANIC); |
bb4f397a | 3524 | else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) |
fd8c37ec | 3525 | set_opt(sb, ERRORS_CONT); |
bb4f397a | 3526 | else |
fd8c37ec | 3527 | set_opt(sb, ERRORS_RO); |
45f1a9c3 DW |
3528 | /* block_validity enabled by default; disable with noblock_validity */ |
3529 | set_opt(sb, BLOCK_VALIDITY); | |
8b67f04a | 3530 | if (def_mount_opts & EXT4_DEFM_DISCARD) |
fd8c37ec | 3531 | set_opt(sb, DISCARD); |
ac27a0ec | 3532 | |
08cefc7a EB |
3533 | sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); |
3534 | sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); | |
30773840 TT |
3535 | sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; |
3536 | sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; | |
3537 | sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; | |
ac27a0ec | 3538 | |
8b67f04a | 3539 | if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) |
fd8c37ec | 3540 | set_opt(sb, BARRIER); |
ac27a0ec | 3541 | |
dd919b98 AK |
3542 | /* |
3543 | * enable delayed allocation by default | |
3544 | * Use -o nodelalloc to turn it off | |
3545 | */ | |
bc0b75f7 | 3546 | if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && |
8b67f04a | 3547 | ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) |
fd8c37ec | 3548 | set_opt(sb, DELALLOC); |
dd919b98 | 3549 | |
51ce6511 LC |
3550 | /* |
3551 | * set default s_li_wait_mult for lazyinit, for the case there is | |
3552 | * no mount option specified. | |
3553 | */ | |
3554 | sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; | |
3555 | ||
5aee0f8a TT |
3556 | if (sbi->s_es->s_mount_opts[0]) { |
3557 | char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts, | |
3558 | sizeof(sbi->s_es->s_mount_opts), | |
3559 | GFP_KERNEL); | |
3560 | if (!s_mount_opts) | |
3561 | goto failed_mount; | |
3562 | if (!parse_options(s_mount_opts, sb, &journal_devnum, | |
3563 | &journal_ioprio, 0)) { | |
3564 | ext4_msg(sb, KERN_WARNING, | |
3565 | "failed to parse options in superblock: %s", | |
3566 | s_mount_opts); | |
3567 | } | |
3568 | kfree(s_mount_opts); | |
8b67f04a | 3569 | } |
5a916be1 | 3570 | sbi->s_def_mount_opt = sbi->s_mount_opt; |
b3881f74 | 3571 | if (!parse_options((char *) data, sb, &journal_devnum, |
661aa520 | 3572 | &journal_ioprio, 0)) |
ac27a0ec DK |
3573 | goto failed_mount; |
3574 | ||
56889787 TT |
3575 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { |
3576 | printk_once(KERN_WARNING "EXT4-fs: Warning: mounting " | |
3577 | "with data=journal disables delayed " | |
3578 | "allocation and O_DIRECT support!\n"); | |
3579 | if (test_opt2(sb, EXPLICIT_DELALLOC)) { | |
3580 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
3581 | "both data=journal and delalloc"); | |
3582 | goto failed_mount; | |
3583 | } | |
3584 | if (test_opt(sb, DIOREAD_NOLOCK)) { | |
3585 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
6ae6514b | 3586 | "both data=journal and dioread_nolock"); |
56889787 TT |
3587 | goto failed_mount; |
3588 | } | |
923ae0ff RZ |
3589 | if (test_opt(sb, DAX)) { |
3590 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
3591 | "both data=journal and dax"); | |
3592 | goto failed_mount; | |
3593 | } | |
73b92a2a SK |
3594 | if (ext4_has_feature_encrypt(sb)) { |
3595 | ext4_msg(sb, KERN_WARNING, | |
3596 | "encrypted files will use data=ordered " | |
3597 | "instead of data journaling mode"); | |
3598 | } | |
56889787 TT |
3599 | if (test_opt(sb, DELALLOC)) |
3600 | clear_opt(sb, DELALLOC); | |
001e4a87 TH |
3601 | } else { |
3602 | sb->s_iflags |= SB_I_CGROUPWB; | |
56889787 TT |
3603 | } |
3604 | ||
ac27a0ec | 3605 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | |
482a7425 | 3606 | (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); |
ac27a0ec | 3607 | |
617ba13b | 3608 | if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && |
e2b911c5 DW |
3609 | (ext4_has_compat_features(sb) || |
3610 | ext4_has_ro_compat_features(sb) || | |
3611 | ext4_has_incompat_features(sb))) | |
b31e1552 ES |
3612 | ext4_msg(sb, KERN_WARNING, |
3613 | "feature flags set on rev 0 fs, " | |
3614 | "running e2fsck is recommended"); | |
469108ff | 3615 | |
ed3654eb TT |
3616 | if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { |
3617 | set_opt2(sb, HURD_COMPAT); | |
e2b911c5 | 3618 | if (ext4_has_feature_64bit(sb)) { |
ed3654eb TT |
3619 | ext4_msg(sb, KERN_ERR, |
3620 | "The Hurd can't support 64-bit file systems"); | |
3621 | goto failed_mount; | |
3622 | } | |
dec214d0 TE |
3623 | |
3624 | /* | |
3625 | * ea_inode feature uses l_i_version field which is not | |
3626 | * available in HURD_COMPAT mode. | |
3627 | */ | |
3628 | if (ext4_has_feature_ea_inode(sb)) { | |
3629 | ext4_msg(sb, KERN_ERR, | |
3630 | "ea_inode feature is not supported for Hurd"); | |
3631 | goto failed_mount; | |
3632 | } | |
ed3654eb TT |
3633 | } |
3634 | ||
2035e776 TT |
3635 | if (IS_EXT2_SB(sb)) { |
3636 | if (ext2_feature_set_ok(sb)) | |
3637 | ext4_msg(sb, KERN_INFO, "mounting ext2 file system " | |
3638 | "using the ext4 subsystem"); | |
3639 | else { | |
3640 | ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " | |
3641 | "to feature incompatibilities"); | |
3642 | goto failed_mount; | |
3643 | } | |
3644 | } | |
3645 | ||
3646 | if (IS_EXT3_SB(sb)) { | |
3647 | if (ext3_feature_set_ok(sb)) | |
3648 | ext4_msg(sb, KERN_INFO, "mounting ext3 file system " | |
3649 | "using the ext4 subsystem"); | |
3650 | else { | |
3651 | ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " | |
3652 | "to feature incompatibilities"); | |
3653 | goto failed_mount; | |
3654 | } | |
3655 | } | |
3656 | ||
ac27a0ec DK |
3657 | /* |
3658 | * Check feature flags regardless of the revision level, since we | |
3659 | * previously didn't change the revision level when setting the flags, | |
3660 | * so there is a chance incompat flags are set on a rev 0 filesystem. | |
3661 | */ | |
bc98a42c | 3662 | if (!ext4_feature_set_ok(sb, (sb_rdonly(sb)))) |
ac27a0ec | 3663 | goto failed_mount; |
a13fb1a4 | 3664 | |
261cb20c | 3665 | blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); |
617ba13b MC |
3666 | if (blocksize < EXT4_MIN_BLOCK_SIZE || |
3667 | blocksize > EXT4_MAX_BLOCK_SIZE) { | |
b31e1552 | 3668 | ext4_msg(sb, KERN_ERR, |
8cdf3372 TT |
3669 | "Unsupported filesystem blocksize %d (%d log_block_size)", |
3670 | blocksize, le32_to_cpu(es->s_log_block_size)); | |
3671 | goto failed_mount; | |
3672 | } | |
3673 | if (le32_to_cpu(es->s_log_block_size) > | |
3674 | (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { | |
3675 | ext4_msg(sb, KERN_ERR, | |
3676 | "Invalid log block size: %u", | |
3677 | le32_to_cpu(es->s_log_block_size)); | |
ac27a0ec DK |
3678 | goto failed_mount; |
3679 | } | |
3680 | ||
5b9554dc TT |
3681 | if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) { |
3682 | ext4_msg(sb, KERN_ERR, | |
3683 | "Number of reserved GDT blocks insanely large: %d", | |
3684 | le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks)); | |
3685 | goto failed_mount; | |
3686 | } | |
3687 | ||
923ae0ff | 3688 | if (sbi->s_mount_opt & EXT4_MOUNT_DAX) { |
87eefeb4 TK |
3689 | err = bdev_dax_supported(sb, blocksize); |
3690 | if (err) | |
923ae0ff | 3691 | goto failed_mount; |
923ae0ff RZ |
3692 | } |
3693 | ||
e2b911c5 | 3694 | if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { |
6ddb2447 TT |
3695 | ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", |
3696 | es->s_encryption_level); | |
3697 | goto failed_mount; | |
3698 | } | |
3699 | ||
ac27a0ec | 3700 | if (sb->s_blocksize != blocksize) { |
ce40733c AK |
3701 | /* Validate the filesystem blocksize */ |
3702 | if (!sb_set_blocksize(sb, blocksize)) { | |
b31e1552 | 3703 | ext4_msg(sb, KERN_ERR, "bad block size %d", |
ce40733c | 3704 | blocksize); |
ac27a0ec DK |
3705 | goto failed_mount; |
3706 | } | |
3707 | ||
2b2d6d01 | 3708 | brelse(bh); |
70bbb3e0 AM |
3709 | logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; |
3710 | offset = do_div(logical_sb_block, blocksize); | |
a8ac900b | 3711 | bh = sb_bread_unmovable(sb, logical_sb_block); |
ac27a0ec | 3712 | if (!bh) { |
b31e1552 ES |
3713 | ext4_msg(sb, KERN_ERR, |
3714 | "Can't read superblock on 2nd try"); | |
ac27a0ec DK |
3715 | goto failed_mount; |
3716 | } | |
2716b802 | 3717 | es = (struct ext4_super_block *)(bh->b_data + offset); |
ac27a0ec | 3718 | sbi->s_es = es; |
617ba13b | 3719 | if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { |
b31e1552 ES |
3720 | ext4_msg(sb, KERN_ERR, |
3721 | "Magic mismatch, very weird!"); | |
ac27a0ec DK |
3722 | goto failed_mount; |
3723 | } | |
3724 | } | |
3725 | ||
e2b911c5 | 3726 | has_huge_files = ext4_has_feature_huge_file(sb); |
f287a1a5 TT |
3727 | sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, |
3728 | has_huge_files); | |
3729 | sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); | |
ac27a0ec | 3730 | |
617ba13b MC |
3731 | if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { |
3732 | sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; | |
3733 | sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; | |
ac27a0ec DK |
3734 | } else { |
3735 | sbi->s_inode_size = le16_to_cpu(es->s_inode_size); | |
3736 | sbi->s_first_ino = le32_to_cpu(es->s_first_ino); | |
617ba13b | 3737 | if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || |
1330593e | 3738 | (!is_power_of_2(sbi->s_inode_size)) || |
ac27a0ec | 3739 | (sbi->s_inode_size > blocksize)) { |
b31e1552 ES |
3740 | ext4_msg(sb, KERN_ERR, |
3741 | "unsupported inode size: %d", | |
2b2d6d01 | 3742 | sbi->s_inode_size); |
ac27a0ec DK |
3743 | goto failed_mount; |
3744 | } | |
ef7f3835 KS |
3745 | if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) |
3746 | sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2); | |
ac27a0ec | 3747 | } |
0b8e58a1 | 3748 | |
0d1ee42f | 3749 | sbi->s_desc_size = le16_to_cpu(es->s_desc_size); |
e2b911c5 | 3750 | if (ext4_has_feature_64bit(sb)) { |
8fadc143 | 3751 | if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || |
0d1ee42f | 3752 | sbi->s_desc_size > EXT4_MAX_DESC_SIZE || |
d8ea6cf8 | 3753 | !is_power_of_2(sbi->s_desc_size)) { |
b31e1552 ES |
3754 | ext4_msg(sb, KERN_ERR, |
3755 | "unsupported descriptor size %lu", | |
0d1ee42f AR |
3756 | sbi->s_desc_size); |
3757 | goto failed_mount; | |
3758 | } | |
3759 | } else | |
3760 | sbi->s_desc_size = EXT4_MIN_DESC_SIZE; | |
0b8e58a1 | 3761 | |
ac27a0ec | 3762 | sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); |
ac27a0ec | 3763 | sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); |
0b8e58a1 | 3764 | |
617ba13b | 3765 | sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); |
ac27a0ec | 3766 | if (sbi->s_inodes_per_block == 0) |
617ba13b | 3767 | goto cantfind_ext4; |
cd6bb35b TT |
3768 | if (sbi->s_inodes_per_group < sbi->s_inodes_per_block || |
3769 | sbi->s_inodes_per_group > blocksize * 8) { | |
3770 | ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n", | |
3771 | sbi->s_blocks_per_group); | |
3772 | goto failed_mount; | |
3773 | } | |
ac27a0ec DK |
3774 | sbi->s_itb_per_group = sbi->s_inodes_per_group / |
3775 | sbi->s_inodes_per_block; | |
0d1ee42f | 3776 | sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); |
ac27a0ec DK |
3777 | sbi->s_sbh = bh; |
3778 | sbi->s_mount_state = le16_to_cpu(es->s_state); | |
e57aa839 FW |
3779 | sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); |
3780 | sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); | |
0b8e58a1 | 3781 | |
2b2d6d01 | 3782 | for (i = 0; i < 4; i++) |
ac27a0ec DK |
3783 | sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); |
3784 | sbi->s_def_hash_version = es->s_def_hash_version; | |
e2b911c5 | 3785 | if (ext4_has_feature_dir_index(sb)) { |
23301410 TT |
3786 | i = le32_to_cpu(es->s_flags); |
3787 | if (i & EXT2_FLAGS_UNSIGNED_HASH) | |
3788 | sbi->s_hash_unsigned = 3; | |
3789 | else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { | |
f99b2589 | 3790 | #ifdef __CHAR_UNSIGNED__ |
bc98a42c | 3791 | if (!sb_rdonly(sb)) |
23301410 TT |
3792 | es->s_flags |= |
3793 | cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); | |
3794 | sbi->s_hash_unsigned = 3; | |
f99b2589 | 3795 | #else |
bc98a42c | 3796 | if (!sb_rdonly(sb)) |
23301410 TT |
3797 | es->s_flags |= |
3798 | cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); | |
f99b2589 | 3799 | #endif |
23301410 | 3800 | } |
f99b2589 | 3801 | } |
ac27a0ec | 3802 | |
281b5995 TT |
3803 | /* Handle clustersize */ |
3804 | clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); | |
e2b911c5 | 3805 | has_bigalloc = ext4_has_feature_bigalloc(sb); |
281b5995 TT |
3806 | if (has_bigalloc) { |
3807 | if (clustersize < blocksize) { | |
3808 | ext4_msg(sb, KERN_ERR, | |
3809 | "cluster size (%d) smaller than " | |
3810 | "block size (%d)", clustersize, blocksize); | |
3811 | goto failed_mount; | |
3812 | } | |
8cdf3372 TT |
3813 | if (le32_to_cpu(es->s_log_cluster_size) > |
3814 | (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { | |
3815 | ext4_msg(sb, KERN_ERR, | |
3816 | "Invalid log cluster size: %u", | |
3817 | le32_to_cpu(es->s_log_cluster_size)); | |
3818 | goto failed_mount; | |
3819 | } | |
281b5995 TT |
3820 | sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - |
3821 | le32_to_cpu(es->s_log_block_size); | |
3822 | sbi->s_clusters_per_group = | |
3823 | le32_to_cpu(es->s_clusters_per_group); | |
3824 | if (sbi->s_clusters_per_group > blocksize * 8) { | |
3825 | ext4_msg(sb, KERN_ERR, | |
3826 | "#clusters per group too big: %lu", | |
3827 | sbi->s_clusters_per_group); | |
3828 | goto failed_mount; | |
3829 | } | |
3830 | if (sbi->s_blocks_per_group != | |
3831 | (sbi->s_clusters_per_group * (clustersize / blocksize))) { | |
3832 | ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " | |
3833 | "clusters per group (%lu) inconsistent", | |
3834 | sbi->s_blocks_per_group, | |
3835 | sbi->s_clusters_per_group); | |
3836 | goto failed_mount; | |
3837 | } | |
3838 | } else { | |
3839 | if (clustersize != blocksize) { | |
3840 | ext4_warning(sb, "fragment/cluster size (%d) != " | |
3841 | "block size (%d)", clustersize, | |
3842 | blocksize); | |
3843 | clustersize = blocksize; | |
3844 | } | |
3845 | if (sbi->s_blocks_per_group > blocksize * 8) { | |
3846 | ext4_msg(sb, KERN_ERR, | |
3847 | "#blocks per group too big: %lu", | |
3848 | sbi->s_blocks_per_group); | |
3849 | goto failed_mount; | |
3850 | } | |
3851 | sbi->s_clusters_per_group = sbi->s_blocks_per_group; | |
3852 | sbi->s_cluster_bits = 0; | |
ac27a0ec | 3853 | } |
281b5995 TT |
3854 | sbi->s_cluster_ratio = clustersize / blocksize; |
3855 | ||
960fd856 TT |
3856 | /* Do we have standard group size of clustersize * 8 blocks ? */ |
3857 | if (sbi->s_blocks_per_group == clustersize << 3) | |
3858 | set_opt2(sb, STD_GROUP_SIZE); | |
3859 | ||
bf43d84b ES |
3860 | /* |
3861 | * Test whether we have more sectors than will fit in sector_t, | |
3862 | * and whether the max offset is addressable by the page cache. | |
3863 | */ | |
5a9ae68a | 3864 | err = generic_check_addressable(sb->s_blocksize_bits, |
30ca22c7 | 3865 | ext4_blocks_count(es)); |
5a9ae68a | 3866 | if (err) { |
b31e1552 | 3867 | ext4_msg(sb, KERN_ERR, "filesystem" |
bf43d84b | 3868 | " too large to mount safely on this system"); |
ac27a0ec | 3869 | if (sizeof(sector_t) < 8) |
90c699a9 | 3870 | ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled"); |
ac27a0ec DK |
3871 | goto failed_mount; |
3872 | } | |
3873 | ||
617ba13b MC |
3874 | if (EXT4_BLOCKS_PER_GROUP(sb) == 0) |
3875 | goto cantfind_ext4; | |
e7c95593 | 3876 | |
0f2ddca6 FTN |
3877 | /* check blocks count against device size */ |
3878 | blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; | |
3879 | if (blocks_count && ext4_blocks_count(es) > blocks_count) { | |
b31e1552 ES |
3880 | ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " |
3881 | "exceeds size of device (%llu blocks)", | |
0f2ddca6 FTN |
3882 | ext4_blocks_count(es), blocks_count); |
3883 | goto failed_mount; | |
3884 | } | |
3885 | ||
0b8e58a1 AD |
3886 | /* |
3887 | * It makes no sense for the first data block to be beyond the end | |
3888 | * of the filesystem. | |
3889 | */ | |
3890 | if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { | |
5635a62b | 3891 | ext4_msg(sb, KERN_WARNING, "bad geometry: first data " |
b31e1552 ES |
3892 | "block %u is beyond end of filesystem (%llu)", |
3893 | le32_to_cpu(es->s_first_data_block), | |
3894 | ext4_blocks_count(es)); | |
e7c95593 ES |
3895 | goto failed_mount; |
3896 | } | |
bd81d8ee LV |
3897 | blocks_count = (ext4_blocks_count(es) - |
3898 | le32_to_cpu(es->s_first_data_block) + | |
3899 | EXT4_BLOCKS_PER_GROUP(sb) - 1); | |
3900 | do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); | |
4ec11028 | 3901 | if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { |
b31e1552 | 3902 | ext4_msg(sb, KERN_WARNING, "groups count too large: %u " |
4ec11028 | 3903 | "(block count %llu, first data block %u, " |
b31e1552 | 3904 | "blocks per group %lu)", sbi->s_groups_count, |
4ec11028 TT |
3905 | ext4_blocks_count(es), |
3906 | le32_to_cpu(es->s_first_data_block), | |
3907 | EXT4_BLOCKS_PER_GROUP(sb)); | |
3908 | goto failed_mount; | |
3909 | } | |
bd81d8ee | 3910 | sbi->s_groups_count = blocks_count; |
fb0a387d ES |
3911 | sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, |
3912 | (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); | |
617ba13b MC |
3913 | db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / |
3914 | EXT4_DESC_PER_BLOCK(sb); | |
3a4b77cd | 3915 | if (ext4_has_feature_meta_bg(sb)) { |
2ba3e6e8 | 3916 | if (le32_to_cpu(es->s_first_meta_bg) > db_count) { |
3a4b77cd EG |
3917 | ext4_msg(sb, KERN_WARNING, |
3918 | "first meta block group too large: %u " | |
3919 | "(group descriptor block count %u)", | |
3920 | le32_to_cpu(es->s_first_meta_bg), db_count); | |
3921 | goto failed_mount; | |
3922 | } | |
3923 | } | |
a7c3e901 | 3924 | sbi->s_group_desc = kvmalloc(db_count * |
f18a5f21 TT |
3925 | sizeof(struct buffer_head *), |
3926 | GFP_KERNEL); | |
ac27a0ec | 3927 | if (sbi->s_group_desc == NULL) { |
b31e1552 | 3928 | ext4_msg(sb, KERN_ERR, "not enough memory"); |
2cde417d | 3929 | ret = -ENOMEM; |
ac27a0ec DK |
3930 | goto failed_mount; |
3931 | } | |
3932 | ||
705895b6 | 3933 | bgl_lock_init(sbi->s_blockgroup_lock); |
ac27a0ec | 3934 | |
85c8f176 AP |
3935 | /* Pre-read the descriptors into the buffer cache */ |
3936 | for (i = 0; i < db_count; i++) { | |
3937 | block = descriptor_loc(sb, logical_sb_block, i); | |
3938 | sb_breadahead(sb, block); | |
3939 | } | |
3940 | ||
ac27a0ec | 3941 | for (i = 0; i < db_count; i++) { |
70bbb3e0 | 3942 | block = descriptor_loc(sb, logical_sb_block, i); |
a8ac900b | 3943 | sbi->s_group_desc[i] = sb_bread_unmovable(sb, block); |
ac27a0ec | 3944 | if (!sbi->s_group_desc[i]) { |
b31e1552 ES |
3945 | ext4_msg(sb, KERN_ERR, |
3946 | "can't read group descriptor %d", i); | |
ac27a0ec DK |
3947 | db_count = i; |
3948 | goto failed_mount2; | |
3949 | } | |
3950 | } | |
829fa70d | 3951 | if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) { |
b31e1552 | 3952 | ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); |
6a797d27 | 3953 | ret = -EFSCORRUPTED; |
f9ae9cf5 | 3954 | goto failed_mount2; |
ac27a0ec | 3955 | } |
772cb7c8 | 3956 | |
f9ae9cf5 | 3957 | sbi->s_gdb_count = db_count; |
ac27a0ec DK |
3958 | get_random_bytes(&sbi->s_next_generation, sizeof(u32)); |
3959 | spin_lock_init(&sbi->s_next_gen_lock); | |
3960 | ||
04ecddb7 JM |
3961 | setup_timer(&sbi->s_err_report, print_daily_error_info, |
3962 | (unsigned long) sb); | |
04496411 | 3963 | |
a75ae78f | 3964 | /* Register extent status tree shrinker */ |
eb68d0e2 | 3965 | if (ext4_es_register_shrinker(sbi)) |
ce7e010a | 3966 | goto failed_mount3; |
ce7e010a | 3967 | |
c9de560d | 3968 | sbi->s_stripe = ext4_get_stripe_size(sbi); |
67a5da56 | 3969 | sbi->s_extent_max_zeroout_kb = 32; |
c9de560d | 3970 | |
f9ae9cf5 TT |
3971 | /* |
3972 | * set up enough so that it can read an inode | |
3973 | */ | |
f6e63f90 | 3974 | sb->s_op = &ext4_sops; |
617ba13b MC |
3975 | sb->s_export_op = &ext4_export_ops; |
3976 | sb->s_xattr = ext4_xattr_handlers; | |
a7550b30 | 3977 | sb->s_cop = &ext4_cryptops; |
ac27a0ec | 3978 | #ifdef CONFIG_QUOTA |
617ba13b | 3979 | sb->dq_op = &ext4_quota_operations; |
e2b911c5 | 3980 | if (ext4_has_feature_quota(sb)) |
1fa5efe3 | 3981 | sb->s_qcop = &dquot_quotactl_sysfile_ops; |
262b4662 JK |
3982 | else |
3983 | sb->s_qcop = &ext4_qctl_operations; | |
689c958c | 3984 | sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; |
ac27a0ec | 3985 | #endif |
85787090 | 3986 | memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); |
f2fa2ffc | 3987 | |
ac27a0ec | 3988 | INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ |
3b9d4ed2 | 3989 | mutex_init(&sbi->s_orphan_lock); |
ac27a0ec DK |
3990 | |
3991 | sb->s_root = NULL; | |
3992 | ||
3993 | needs_recovery = (es->s_last_orphan != 0 || | |
e2b911c5 | 3994 | ext4_has_feature_journal_needs_recovery(sb)); |
ac27a0ec | 3995 | |
bc98a42c | 3996 | if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) |
c5e06d10 | 3997 | if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) |
50460fe8 | 3998 | goto failed_mount3a; |
c5e06d10 | 3999 | |
ac27a0ec DK |
4000 | /* |
4001 | * The first inode we look at is the journal inode. Don't try | |
4002 | * root first: it may be modified in the journal! | |
4003 | */ | |
e2b911c5 | 4004 | if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { |
4753d8a2 TT |
4005 | err = ext4_load_journal(sb, es, journal_devnum); |
4006 | if (err) | |
50460fe8 | 4007 | goto failed_mount3a; |
bc98a42c | 4008 | } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) && |
e2b911c5 | 4009 | ext4_has_feature_journal_needs_recovery(sb)) { |
b31e1552 ES |
4010 | ext4_msg(sb, KERN_ERR, "required journal recovery " |
4011 | "suppressed and not mounted read-only"); | |
744692dc | 4012 | goto failed_mount_wq; |
ac27a0ec | 4013 | } else { |
1e381f60 DM |
4014 | /* Nojournal mode, all journal mount options are illegal */ |
4015 | if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) { | |
4016 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
4017 | "journal_checksum, fs mounted w/o journal"); | |
4018 | goto failed_mount_wq; | |
4019 | } | |
4020 | if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { | |
4021 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
4022 | "journal_async_commit, fs mounted w/o journal"); | |
4023 | goto failed_mount_wq; | |
4024 | } | |
4025 | if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { | |
4026 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
4027 | "commit=%lu, fs mounted w/o journal", | |
4028 | sbi->s_commit_interval / HZ); | |
4029 | goto failed_mount_wq; | |
4030 | } | |
4031 | if (EXT4_MOUNT_DATA_FLAGS & | |
4032 | (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { | |
4033 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
4034 | "data=, fs mounted w/o journal"); | |
4035 | goto failed_mount_wq; | |
4036 | } | |
4037 | sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM; | |
4038 | clear_opt(sb, JOURNAL_CHECKSUM); | |
fd8c37ec | 4039 | clear_opt(sb, DATA_FLAGS); |
0390131b FM |
4040 | sbi->s_journal = NULL; |
4041 | needs_recovery = 0; | |
4042 | goto no_journal; | |
ac27a0ec DK |
4043 | } |
4044 | ||
e2b911c5 | 4045 | if (ext4_has_feature_64bit(sb) && |
eb40a09c JS |
4046 | !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, |
4047 | JBD2_FEATURE_INCOMPAT_64BIT)) { | |
b31e1552 | 4048 | ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); |
744692dc | 4049 | goto failed_mount_wq; |
eb40a09c JS |
4050 | } |
4051 | ||
25ed6e8a DW |
4052 | if (!set_journal_csum_feature_set(sb)) { |
4053 | ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " | |
4054 | "feature set"); | |
4055 | goto failed_mount_wq; | |
d4da6c9c | 4056 | } |
818d276c | 4057 | |
ac27a0ec DK |
4058 | /* We have now updated the journal if required, so we can |
4059 | * validate the data journaling mode. */ | |
4060 | switch (test_opt(sb, DATA_FLAGS)) { | |
4061 | case 0: | |
4062 | /* No mode set, assume a default based on the journal | |
63f57933 AM |
4063 | * capabilities: ORDERED_DATA if the journal can |
4064 | * cope, else JOURNAL_DATA | |
4065 | */ | |
dab291af MC |
4066 | if (jbd2_journal_check_available_features |
4067 | (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) | |
fd8c37ec | 4068 | set_opt(sb, ORDERED_DATA); |
ac27a0ec | 4069 | else |
fd8c37ec | 4070 | set_opt(sb, JOURNAL_DATA); |
ac27a0ec DK |
4071 | break; |
4072 | ||
617ba13b MC |
4073 | case EXT4_MOUNT_ORDERED_DATA: |
4074 | case EXT4_MOUNT_WRITEBACK_DATA: | |
dab291af MC |
4075 | if (!jbd2_journal_check_available_features |
4076 | (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { | |
b31e1552 ES |
4077 | ext4_msg(sb, KERN_ERR, "Journal does not support " |
4078 | "requested data journaling mode"); | |
744692dc | 4079 | goto failed_mount_wq; |
ac27a0ec DK |
4080 | } |
4081 | default: | |
4082 | break; | |
4083 | } | |
ab04df78 JK |
4084 | |
4085 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && | |
4086 | test_opt(sb, JOURNAL_ASYNC_COMMIT)) { | |
4087 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
4088 | "journal_async_commit in data=ordered mode"); | |
4089 | goto failed_mount_wq; | |
4090 | } | |
4091 | ||
b3881f74 | 4092 | set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); |
ac27a0ec | 4093 | |
18aadd47 BJ |
4094 | sbi->s_journal->j_commit_callback = ext4_journal_commit_callback; |
4095 | ||
ce7e010a | 4096 | no_journal: |
cdb7ee4c TE |
4097 | if (!test_opt(sb, NO_MBCACHE)) { |
4098 | sbi->s_ea_block_cache = ext4_xattr_create_cache(); | |
4099 | if (!sbi->s_ea_block_cache) { | |
dec214d0 | 4100 | ext4_msg(sb, KERN_ERR, |
cdb7ee4c | 4101 | "Failed to create ea_block_cache"); |
dec214d0 TE |
4102 | goto failed_mount_wq; |
4103 | } | |
cdb7ee4c TE |
4104 | |
4105 | if (ext4_has_feature_ea_inode(sb)) { | |
4106 | sbi->s_ea_inode_cache = ext4_xattr_create_cache(); | |
4107 | if (!sbi->s_ea_inode_cache) { | |
4108 | ext4_msg(sb, KERN_ERR, | |
4109 | "Failed to create ea_inode_cache"); | |
4110 | goto failed_mount_wq; | |
4111 | } | |
4112 | } | |
9c191f70 M |
4113 | } |
4114 | ||
e2b911c5 | 4115 | if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) && |
09cbfeaf | 4116 | (blocksize != PAGE_SIZE)) { |
1cb767cd TT |
4117 | ext4_msg(sb, KERN_ERR, |
4118 | "Unsupported blocksize for fs encryption"); | |
4119 | goto failed_mount_wq; | |
4120 | } | |
4121 | ||
bc98a42c | 4122 | if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) && |
e2b911c5 DW |
4123 | !ext4_has_feature_encrypt(sb)) { |
4124 | ext4_set_feature_encrypt(sb); | |
6ddb2447 TT |
4125 | ext4_commit_super(sb, 1); |
4126 | } | |
4127 | ||
952fc18e TT |
4128 | /* |
4129 | * Get the # of file system overhead blocks from the | |
4130 | * superblock if present. | |
4131 | */ | |
4132 | if (es->s_overhead_clusters) | |
4133 | sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); | |
4134 | else { | |
07aa2ea1 LC |
4135 | err = ext4_calculate_overhead(sb); |
4136 | if (err) | |
952fc18e TT |
4137 | goto failed_mount_wq; |
4138 | } | |
4139 | ||
fd89d5f2 TH |
4140 | /* |
4141 | * The maximum number of concurrent works can be high and | |
4142 | * concurrency isn't really necessary. Limit it to 1. | |
4143 | */ | |
2e8fa54e JK |
4144 | EXT4_SB(sb)->rsv_conversion_wq = |
4145 | alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); | |
4146 | if (!EXT4_SB(sb)->rsv_conversion_wq) { | |
4147 | printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); | |
07aa2ea1 | 4148 | ret = -ENOMEM; |
2e8fa54e JK |
4149 | goto failed_mount4; |
4150 | } | |
4151 | ||
ac27a0ec | 4152 | /* |
dab291af | 4153 | * The jbd2_journal_load will have done any necessary log recovery, |
ac27a0ec DK |
4154 | * so we can safely mount the rest of the filesystem now. |
4155 | */ | |
4156 | ||
1d1fe1ee DH |
4157 | root = ext4_iget(sb, EXT4_ROOT_INO); |
4158 | if (IS_ERR(root)) { | |
b31e1552 | 4159 | ext4_msg(sb, KERN_ERR, "get root inode failed"); |
1d1fe1ee | 4160 | ret = PTR_ERR(root); |
32a9bb57 | 4161 | root = NULL; |
ac27a0ec DK |
4162 | goto failed_mount4; |
4163 | } | |
4164 | if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { | |
b31e1552 | 4165 | ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); |
94bf608a | 4166 | iput(root); |
ac27a0ec DK |
4167 | goto failed_mount4; |
4168 | } | |
48fde701 | 4169 | sb->s_root = d_make_root(root); |
1d1fe1ee | 4170 | if (!sb->s_root) { |
b31e1552 | 4171 | ext4_msg(sb, KERN_ERR, "get root dentry failed"); |
1d1fe1ee DH |
4172 | ret = -ENOMEM; |
4173 | goto failed_mount4; | |
4174 | } | |
ac27a0ec | 4175 | |
bc98a42c | 4176 | if (ext4_setup_super(sb, es, sb_rdonly(sb))) |
7e84b621 | 4177 | sb->s_flags |= MS_RDONLY; |
ef7f3835 KS |
4178 | |
4179 | /* determine the minimum size of new large inodes, if present */ | |
670e9875 TT |
4180 | if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE && |
4181 | sbi->s_want_extra_isize == 0) { | |
ef7f3835 KS |
4182 | sbi->s_want_extra_isize = sizeof(struct ext4_inode) - |
4183 | EXT4_GOOD_OLD_INODE_SIZE; | |
e2b911c5 | 4184 | if (ext4_has_feature_extra_isize(sb)) { |
ef7f3835 KS |
4185 | if (sbi->s_want_extra_isize < |
4186 | le16_to_cpu(es->s_want_extra_isize)) | |
4187 | sbi->s_want_extra_isize = | |
4188 | le16_to_cpu(es->s_want_extra_isize); | |
4189 | if (sbi->s_want_extra_isize < | |
4190 | le16_to_cpu(es->s_min_extra_isize)) | |
4191 | sbi->s_want_extra_isize = | |
4192 | le16_to_cpu(es->s_min_extra_isize); | |
4193 | } | |
4194 | } | |
4195 | /* Check if enough inode space is available */ | |
4196 | if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize > | |
4197 | sbi->s_inode_size) { | |
4198 | sbi->s_want_extra_isize = sizeof(struct ext4_inode) - | |
4199 | EXT4_GOOD_OLD_INODE_SIZE; | |
b31e1552 ES |
4200 | ext4_msg(sb, KERN_INFO, "required extra inode space not" |
4201 | "available"); | |
ef7f3835 KS |
4202 | } |
4203 | ||
b5799018 | 4204 | ext4_set_resv_clusters(sb); |
27dd4385 | 4205 | |
6fd058f7 TT |
4206 | err = ext4_setup_system_zone(sb); |
4207 | if (err) { | |
b31e1552 | 4208 | ext4_msg(sb, KERN_ERR, "failed to initialize system " |
fbe845dd | 4209 | "zone (%d)", err); |
f9ae9cf5 TT |
4210 | goto failed_mount4a; |
4211 | } | |
4212 | ||
4213 | ext4_ext_init(sb); | |
4214 | err = ext4_mb_init(sb); | |
4215 | if (err) { | |
4216 | ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", | |
4217 | err); | |
dcf2d804 | 4218 | goto failed_mount5; |
c2774d84 AK |
4219 | } |
4220 | ||
d5e03cbb TT |
4221 | block = ext4_count_free_clusters(sb); |
4222 | ext4_free_blocks_count_set(sbi->s_es, | |
4223 | EXT4_C2B(sbi, block)); | |
908c7f19 TH |
4224 | err = percpu_counter_init(&sbi->s_freeclusters_counter, block, |
4225 | GFP_KERNEL); | |
d5e03cbb TT |
4226 | if (!err) { |
4227 | unsigned long freei = ext4_count_free_inodes(sb); | |
4228 | sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); | |
908c7f19 TH |
4229 | err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, |
4230 | GFP_KERNEL); | |
d5e03cbb TT |
4231 | } |
4232 | if (!err) | |
4233 | err = percpu_counter_init(&sbi->s_dirs_counter, | |
908c7f19 | 4234 | ext4_count_dirs(sb), GFP_KERNEL); |
d5e03cbb | 4235 | if (!err) |
908c7f19 TH |
4236 | err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, |
4237 | GFP_KERNEL); | |
c8585c6f DJ |
4238 | if (!err) |
4239 | err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem); | |
4240 | ||
d5e03cbb TT |
4241 | if (err) { |
4242 | ext4_msg(sb, KERN_ERR, "insufficient memory"); | |
4243 | goto failed_mount6; | |
4244 | } | |
4245 | ||
e2b911c5 | 4246 | if (ext4_has_feature_flex_bg(sb)) |
d5e03cbb TT |
4247 | if (!ext4_fill_flex_info(sb)) { |
4248 | ext4_msg(sb, KERN_ERR, | |
4249 | "unable to initialize " | |
4250 | "flex_bg meta info!"); | |
4251 | goto failed_mount6; | |
4252 | } | |
4253 | ||
bfff6873 LC |
4254 | err = ext4_register_li_request(sb, first_not_zeroed); |
4255 | if (err) | |
dcf2d804 | 4256 | goto failed_mount6; |
bfff6873 | 4257 | |
b5799018 | 4258 | err = ext4_register_sysfs(sb); |
dcf2d804 TM |
4259 | if (err) |
4260 | goto failed_mount7; | |
3197ebdb | 4261 | |
9b2ff357 JK |
4262 | #ifdef CONFIG_QUOTA |
4263 | /* Enable quota usage during mount. */ | |
bc98a42c | 4264 | if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) { |
9b2ff357 JK |
4265 | err = ext4_enable_quotas(sb); |
4266 | if (err) | |
4267 | goto failed_mount8; | |
4268 | } | |
4269 | #endif /* CONFIG_QUOTA */ | |
4270 | ||
617ba13b MC |
4271 | EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; |
4272 | ext4_orphan_cleanup(sb, es); | |
4273 | EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; | |
0390131b | 4274 | if (needs_recovery) { |
b31e1552 | 4275 | ext4_msg(sb, KERN_INFO, "recovery complete"); |
0390131b FM |
4276 | ext4_mark_recovery_complete(sb, es); |
4277 | } | |
4278 | if (EXT4_SB(sb)->s_journal) { | |
4279 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) | |
4280 | descr = " journalled data mode"; | |
4281 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) | |
4282 | descr = " ordered data mode"; | |
4283 | else | |
4284 | descr = " writeback data mode"; | |
4285 | } else | |
4286 | descr = "out journal"; | |
4287 | ||
79add3a3 LC |
4288 | if (test_opt(sb, DISCARD)) { |
4289 | struct request_queue *q = bdev_get_queue(sb->s_bdev); | |
4290 | if (!blk_queue_discard(q)) | |
4291 | ext4_msg(sb, KERN_WARNING, | |
4292 | "mounting with \"discard\" option, but " | |
4293 | "the device does not support discard"); | |
4294 | } | |
4295 | ||
e294a537 TT |
4296 | if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) |
4297 | ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " | |
5aee0f8a TT |
4298 | "Opts: %.*s%s%s", descr, |
4299 | (int) sizeof(sbi->s_es->s_mount_opts), | |
4300 | sbi->s_es->s_mount_opts, | |
e294a537 | 4301 | *sbi->s_es->s_mount_opts ? "; " : "", orig_data); |
ac27a0ec | 4302 | |
66e61a9e TT |
4303 | if (es->s_error_count) |
4304 | mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ | |
ac27a0ec | 4305 | |
efbed4dc TT |
4306 | /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ |
4307 | ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); | |
4308 | ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); | |
4309 | ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); | |
4310 | ||
d4c402d9 | 4311 | kfree(orig_data); |
ac27a0ec DK |
4312 | return 0; |
4313 | ||
617ba13b | 4314 | cantfind_ext4: |
ac27a0ec | 4315 | if (!silent) |
b31e1552 | 4316 | ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); |
ac27a0ec DK |
4317 | goto failed_mount; |
4318 | ||
72ba7450 TT |
4319 | #ifdef CONFIG_QUOTA |
4320 | failed_mount8: | |
ebd173be | 4321 | ext4_unregister_sysfs(sb); |
72ba7450 | 4322 | #endif |
dcf2d804 TM |
4323 | failed_mount7: |
4324 | ext4_unregister_li_request(sb); | |
4325 | failed_mount6: | |
f9ae9cf5 | 4326 | ext4_mb_release(sb); |
d5e03cbb | 4327 | if (sbi->s_flex_groups) |
b93b41d4 | 4328 | kvfree(sbi->s_flex_groups); |
d5e03cbb TT |
4329 | percpu_counter_destroy(&sbi->s_freeclusters_counter); |
4330 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
4331 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
4332 | percpu_counter_destroy(&sbi->s_dirtyclusters_counter); | |
00764937 | 4333 | failed_mount5: |
f9ae9cf5 TT |
4334 | ext4_ext_release(sb); |
4335 | ext4_release_system_zone(sb); | |
4336 | failed_mount4a: | |
94bf608a | 4337 | dput(sb->s_root); |
32a9bb57 | 4338 | sb->s_root = NULL; |
94bf608a | 4339 | failed_mount4: |
b31e1552 | 4340 | ext4_msg(sb, KERN_ERR, "mount failed"); |
2e8fa54e JK |
4341 | if (EXT4_SB(sb)->rsv_conversion_wq) |
4342 | destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); | |
4c0425ff | 4343 | failed_mount_wq: |
dec214d0 TE |
4344 | if (sbi->s_ea_inode_cache) { |
4345 | ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); | |
4346 | sbi->s_ea_inode_cache = NULL; | |
4347 | } | |
47387409 TE |
4348 | if (sbi->s_ea_block_cache) { |
4349 | ext4_xattr_destroy_cache(sbi->s_ea_block_cache); | |
4350 | sbi->s_ea_block_cache = NULL; | |
82939d79 | 4351 | } |
0390131b FM |
4352 | if (sbi->s_journal) { |
4353 | jbd2_journal_destroy(sbi->s_journal); | |
4354 | sbi->s_journal = NULL; | |
4355 | } | |
50460fe8 | 4356 | failed_mount3a: |
d3922a77 | 4357 | ext4_es_unregister_shrinker(sbi); |
eb68d0e2 | 4358 | failed_mount3: |
9105bb14 | 4359 | del_timer_sync(&sbi->s_err_report); |
c5e06d10 JL |
4360 | if (sbi->s_mmp_tsk) |
4361 | kthread_stop(sbi->s_mmp_tsk); | |
ac27a0ec DK |
4362 | failed_mount2: |
4363 | for (i = 0; i < db_count; i++) | |
4364 | brelse(sbi->s_group_desc[i]); | |
b93b41d4 | 4365 | kvfree(sbi->s_group_desc); |
ac27a0ec | 4366 | failed_mount: |
0441984a DW |
4367 | if (sbi->s_chksum_driver) |
4368 | crypto_free_shash(sbi->s_chksum_driver); | |
ac27a0ec | 4369 | #ifdef CONFIG_QUOTA |
a2d4a646 | 4370 | for (i = 0; i < EXT4_MAXQUOTAS; i++) |
ac27a0ec DK |
4371 | kfree(sbi->s_qf_names[i]); |
4372 | #endif | |
617ba13b | 4373 | ext4_blkdev_remove(sbi); |
ac27a0ec DK |
4374 | brelse(bh); |
4375 | out_fail: | |
4376 | sb->s_fs_info = NULL; | |
f6830165 | 4377 | kfree(sbi->s_blockgroup_lock); |
5aee0f8a | 4378 | out_free_base: |
ac27a0ec | 4379 | kfree(sbi); |
d4c402d9 | 4380 | kfree(orig_data); |
07aa2ea1 | 4381 | return err ? err : ret; |
ac27a0ec DK |
4382 | } |
4383 | ||
4384 | /* | |
4385 | * Setup any per-fs journal parameters now. We'll do this both on | |
4386 | * initial mount, once the journal has been initialised but before we've | |
4387 | * done any recovery; and again on any subsequent remount. | |
4388 | */ | |
617ba13b | 4389 | static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) |
ac27a0ec | 4390 | { |
617ba13b | 4391 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 4392 | |
30773840 TT |
4393 | journal->j_commit_interval = sbi->s_commit_interval; |
4394 | journal->j_min_batch_time = sbi->s_min_batch_time; | |
4395 | journal->j_max_batch_time = sbi->s_max_batch_time; | |
ac27a0ec | 4396 | |
a931da6a | 4397 | write_lock(&journal->j_state_lock); |
ac27a0ec | 4398 | if (test_opt(sb, BARRIER)) |
dab291af | 4399 | journal->j_flags |= JBD2_BARRIER; |
ac27a0ec | 4400 | else |
dab291af | 4401 | journal->j_flags &= ~JBD2_BARRIER; |
5bf5683a HK |
4402 | if (test_opt(sb, DATA_ERR_ABORT)) |
4403 | journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; | |
4404 | else | |
4405 | journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; | |
a931da6a | 4406 | write_unlock(&journal->j_state_lock); |
ac27a0ec DK |
4407 | } |
4408 | ||
c6cb7e77 EW |
4409 | static struct inode *ext4_get_journal_inode(struct super_block *sb, |
4410 | unsigned int journal_inum) | |
ac27a0ec DK |
4411 | { |
4412 | struct inode *journal_inode; | |
ac27a0ec | 4413 | |
c6cb7e77 EW |
4414 | /* |
4415 | * Test for the existence of a valid inode on disk. Bad things | |
4416 | * happen if we iget() an unused inode, as the subsequent iput() | |
4417 | * will try to delete it. | |
4418 | */ | |
1d1fe1ee DH |
4419 | journal_inode = ext4_iget(sb, journal_inum); |
4420 | if (IS_ERR(journal_inode)) { | |
b31e1552 | 4421 | ext4_msg(sb, KERN_ERR, "no journal found"); |
ac27a0ec DK |
4422 | return NULL; |
4423 | } | |
4424 | if (!journal_inode->i_nlink) { | |
4425 | make_bad_inode(journal_inode); | |
4426 | iput(journal_inode); | |
b31e1552 | 4427 | ext4_msg(sb, KERN_ERR, "journal inode is deleted"); |
ac27a0ec DK |
4428 | return NULL; |
4429 | } | |
4430 | ||
e5f8eab8 | 4431 | jbd_debug(2, "Journal inode found at %p: %lld bytes\n", |
ac27a0ec | 4432 | journal_inode, journal_inode->i_size); |
1d1fe1ee | 4433 | if (!S_ISREG(journal_inode->i_mode)) { |
b31e1552 | 4434 | ext4_msg(sb, KERN_ERR, "invalid journal inode"); |
ac27a0ec DK |
4435 | iput(journal_inode); |
4436 | return NULL; | |
4437 | } | |
c6cb7e77 EW |
4438 | return journal_inode; |
4439 | } | |
4440 | ||
4441 | static journal_t *ext4_get_journal(struct super_block *sb, | |
4442 | unsigned int journal_inum) | |
4443 | { | |
4444 | struct inode *journal_inode; | |
4445 | journal_t *journal; | |
4446 | ||
4447 | BUG_ON(!ext4_has_feature_journal(sb)); | |
4448 | ||
4449 | journal_inode = ext4_get_journal_inode(sb, journal_inum); | |
4450 | if (!journal_inode) | |
4451 | return NULL; | |
ac27a0ec | 4452 | |
dab291af | 4453 | journal = jbd2_journal_init_inode(journal_inode); |
ac27a0ec | 4454 | if (!journal) { |
b31e1552 | 4455 | ext4_msg(sb, KERN_ERR, "Could not load journal inode"); |
ac27a0ec DK |
4456 | iput(journal_inode); |
4457 | return NULL; | |
4458 | } | |
4459 | journal->j_private = sb; | |
617ba13b | 4460 | ext4_init_journal_params(sb, journal); |
ac27a0ec DK |
4461 | return journal; |
4462 | } | |
4463 | ||
617ba13b | 4464 | static journal_t *ext4_get_dev_journal(struct super_block *sb, |
ac27a0ec DK |
4465 | dev_t j_dev) |
4466 | { | |
2b2d6d01 | 4467 | struct buffer_head *bh; |
ac27a0ec | 4468 | journal_t *journal; |
617ba13b MC |
4469 | ext4_fsblk_t start; |
4470 | ext4_fsblk_t len; | |
ac27a0ec | 4471 | int hblock, blocksize; |
617ba13b | 4472 | ext4_fsblk_t sb_block; |
ac27a0ec | 4473 | unsigned long offset; |
2b2d6d01 | 4474 | struct ext4_super_block *es; |
ac27a0ec DK |
4475 | struct block_device *bdev; |
4476 | ||
e2b911c5 | 4477 | BUG_ON(!ext4_has_feature_journal(sb)); |
0390131b | 4478 | |
b31e1552 | 4479 | bdev = ext4_blkdev_get(j_dev, sb); |
ac27a0ec DK |
4480 | if (bdev == NULL) |
4481 | return NULL; | |
4482 | ||
ac27a0ec | 4483 | blocksize = sb->s_blocksize; |
e1defc4f | 4484 | hblock = bdev_logical_block_size(bdev); |
ac27a0ec | 4485 | if (blocksize < hblock) { |
b31e1552 ES |
4486 | ext4_msg(sb, KERN_ERR, |
4487 | "blocksize too small for journal device"); | |
ac27a0ec DK |
4488 | goto out_bdev; |
4489 | } | |
4490 | ||
617ba13b MC |
4491 | sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; |
4492 | offset = EXT4_MIN_BLOCK_SIZE % blocksize; | |
ac27a0ec DK |
4493 | set_blocksize(bdev, blocksize); |
4494 | if (!(bh = __bread(bdev, sb_block, blocksize))) { | |
b31e1552 ES |
4495 | ext4_msg(sb, KERN_ERR, "couldn't read superblock of " |
4496 | "external journal"); | |
ac27a0ec DK |
4497 | goto out_bdev; |
4498 | } | |
4499 | ||
2716b802 | 4500 | es = (struct ext4_super_block *) (bh->b_data + offset); |
617ba13b | 4501 | if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || |
ac27a0ec | 4502 | !(le32_to_cpu(es->s_feature_incompat) & |
617ba13b | 4503 | EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { |
b31e1552 ES |
4504 | ext4_msg(sb, KERN_ERR, "external journal has " |
4505 | "bad superblock"); | |
ac27a0ec DK |
4506 | brelse(bh); |
4507 | goto out_bdev; | |
4508 | } | |
4509 | ||
df4763be DW |
4510 | if ((le32_to_cpu(es->s_feature_ro_compat) & |
4511 | EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && | |
4512 | es->s_checksum != ext4_superblock_csum(sb, es)) { | |
4513 | ext4_msg(sb, KERN_ERR, "external journal has " | |
4514 | "corrupt superblock"); | |
4515 | brelse(bh); | |
4516 | goto out_bdev; | |
4517 | } | |
4518 | ||
617ba13b | 4519 | if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { |
b31e1552 | 4520 | ext4_msg(sb, KERN_ERR, "journal UUID does not match"); |
ac27a0ec DK |
4521 | brelse(bh); |
4522 | goto out_bdev; | |
4523 | } | |
4524 | ||
bd81d8ee | 4525 | len = ext4_blocks_count(es); |
ac27a0ec DK |
4526 | start = sb_block + 1; |
4527 | brelse(bh); /* we're done with the superblock */ | |
4528 | ||
dab291af | 4529 | journal = jbd2_journal_init_dev(bdev, sb->s_bdev, |
ac27a0ec DK |
4530 | start, len, blocksize); |
4531 | if (!journal) { | |
b31e1552 | 4532 | ext4_msg(sb, KERN_ERR, "failed to create device journal"); |
ac27a0ec DK |
4533 | goto out_bdev; |
4534 | } | |
4535 | journal->j_private = sb; | |
dfec8a14 | 4536 | ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer); |
ac27a0ec DK |
4537 | wait_on_buffer(journal->j_sb_buffer); |
4538 | if (!buffer_uptodate(journal->j_sb_buffer)) { | |
b31e1552 | 4539 | ext4_msg(sb, KERN_ERR, "I/O error on journal device"); |
ac27a0ec DK |
4540 | goto out_journal; |
4541 | } | |
4542 | if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { | |
b31e1552 ES |
4543 | ext4_msg(sb, KERN_ERR, "External journal has more than one " |
4544 | "user (unsupported) - %d", | |
ac27a0ec DK |
4545 | be32_to_cpu(journal->j_superblock->s_nr_users)); |
4546 | goto out_journal; | |
4547 | } | |
617ba13b MC |
4548 | EXT4_SB(sb)->journal_bdev = bdev; |
4549 | ext4_init_journal_params(sb, journal); | |
ac27a0ec | 4550 | return journal; |
0b8e58a1 | 4551 | |
ac27a0ec | 4552 | out_journal: |
dab291af | 4553 | jbd2_journal_destroy(journal); |
ac27a0ec | 4554 | out_bdev: |
617ba13b | 4555 | ext4_blkdev_put(bdev); |
ac27a0ec DK |
4556 | return NULL; |
4557 | } | |
4558 | ||
617ba13b MC |
4559 | static int ext4_load_journal(struct super_block *sb, |
4560 | struct ext4_super_block *es, | |
ac27a0ec DK |
4561 | unsigned long journal_devnum) |
4562 | { | |
4563 | journal_t *journal; | |
4564 | unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); | |
4565 | dev_t journal_dev; | |
4566 | int err = 0; | |
4567 | int really_read_only; | |
4568 | ||
e2b911c5 | 4569 | BUG_ON(!ext4_has_feature_journal(sb)); |
0390131b | 4570 | |
ac27a0ec DK |
4571 | if (journal_devnum && |
4572 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { | |
b31e1552 ES |
4573 | ext4_msg(sb, KERN_INFO, "external journal device major/minor " |
4574 | "numbers have changed"); | |
ac27a0ec DK |
4575 | journal_dev = new_decode_dev(journal_devnum); |
4576 | } else | |
4577 | journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); | |
4578 | ||
4579 | really_read_only = bdev_read_only(sb->s_bdev); | |
4580 | ||
4581 | /* | |
4582 | * Are we loading a blank journal or performing recovery after a | |
4583 | * crash? For recovery, we need to check in advance whether we | |
4584 | * can get read-write access to the device. | |
4585 | */ | |
e2b911c5 | 4586 | if (ext4_has_feature_journal_needs_recovery(sb)) { |
bc98a42c | 4587 | if (sb_rdonly(sb)) { |
b31e1552 ES |
4588 | ext4_msg(sb, KERN_INFO, "INFO: recovery " |
4589 | "required on readonly filesystem"); | |
ac27a0ec | 4590 | if (really_read_only) { |
b31e1552 ES |
4591 | ext4_msg(sb, KERN_ERR, "write access " |
4592 | "unavailable, cannot proceed"); | |
ac27a0ec DK |
4593 | return -EROFS; |
4594 | } | |
b31e1552 ES |
4595 | ext4_msg(sb, KERN_INFO, "write access will " |
4596 | "be enabled during recovery"); | |
ac27a0ec DK |
4597 | } |
4598 | } | |
4599 | ||
4600 | if (journal_inum && journal_dev) { | |
b31e1552 ES |
4601 | ext4_msg(sb, KERN_ERR, "filesystem has both journal " |
4602 | "and inode journals!"); | |
ac27a0ec DK |
4603 | return -EINVAL; |
4604 | } | |
4605 | ||
4606 | if (journal_inum) { | |
617ba13b | 4607 | if (!(journal = ext4_get_journal(sb, journal_inum))) |
ac27a0ec DK |
4608 | return -EINVAL; |
4609 | } else { | |
617ba13b | 4610 | if (!(journal = ext4_get_dev_journal(sb, journal_dev))) |
ac27a0ec DK |
4611 | return -EINVAL; |
4612 | } | |
4613 | ||
90576c0b | 4614 | if (!(journal->j_flags & JBD2_BARRIER)) |
b31e1552 | 4615 | ext4_msg(sb, KERN_INFO, "barriers disabled"); |
4776004f | 4616 | |
e2b911c5 | 4617 | if (!ext4_has_feature_journal_needs_recovery(sb)) |
dab291af | 4618 | err = jbd2_journal_wipe(journal, !really_read_only); |
1c13d5c0 TT |
4619 | if (!err) { |
4620 | char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); | |
4621 | if (save) | |
4622 | memcpy(save, ((char *) es) + | |
4623 | EXT4_S_ERR_START, EXT4_S_ERR_LEN); | |
dab291af | 4624 | err = jbd2_journal_load(journal); |
1c13d5c0 TT |
4625 | if (save) |
4626 | memcpy(((char *) es) + EXT4_S_ERR_START, | |
4627 | save, EXT4_S_ERR_LEN); | |
4628 | kfree(save); | |
4629 | } | |
ac27a0ec DK |
4630 | |
4631 | if (err) { | |
b31e1552 | 4632 | ext4_msg(sb, KERN_ERR, "error loading journal"); |
dab291af | 4633 | jbd2_journal_destroy(journal); |
ac27a0ec DK |
4634 | return err; |
4635 | } | |
4636 | ||
617ba13b MC |
4637 | EXT4_SB(sb)->s_journal = journal; |
4638 | ext4_clear_journal_err(sb, es); | |
ac27a0ec | 4639 | |
c41303ce | 4640 | if (!really_read_only && journal_devnum && |
ac27a0ec DK |
4641 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { |
4642 | es->s_journal_dev = cpu_to_le32(journal_devnum); | |
ac27a0ec DK |
4643 | |
4644 | /* Make sure we flush the recovery flag to disk. */ | |
e2d67052 | 4645 | ext4_commit_super(sb, 1); |
ac27a0ec DK |
4646 | } |
4647 | ||
4648 | return 0; | |
4649 | } | |
4650 | ||
e2d67052 | 4651 | static int ext4_commit_super(struct super_block *sb, int sync) |
ac27a0ec | 4652 | { |
e2d67052 | 4653 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
617ba13b | 4654 | struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; |
c4be0c1d | 4655 | int error = 0; |
ac27a0ec | 4656 | |
bdfe0cbd | 4657 | if (!sbh || block_device_ejected(sb)) |
c4be0c1d | 4658 | return error; |
71290b36 TT |
4659 | /* |
4660 | * If the file system is mounted read-only, don't update the | |
4661 | * superblock write time. This avoids updating the superblock | |
4662 | * write time when we are mounting the root file system | |
4663 | * read/only but we need to replay the journal; at that point, | |
4664 | * for people who are east of GMT and who make their clock | |
4665 | * tick in localtime for Windows bug-for-bug compatibility, | |
4666 | * the clock is set in the future, and this will cause e2fsck | |
4667 | * to complain and force a full file system check. | |
4668 | */ | |
4669 | if (!(sb->s_flags & MS_RDONLY)) | |
4670 | es->s_wtime = cpu_to_le32(get_seconds()); | |
f613dfcb TT |
4671 | if (sb->s_bdev->bd_part) |
4672 | es->s_kbytes_written = | |
4673 | cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + | |
afc32f7e TT |
4674 | ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - |
4675 | EXT4_SB(sb)->s_sectors_written_start) >> 1)); | |
f613dfcb TT |
4676 | else |
4677 | es->s_kbytes_written = | |
4678 | cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); | |
d5e03cbb TT |
4679 | if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter)) |
4680 | ext4_free_blocks_count_set(es, | |
57042651 TT |
4681 | EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive( |
4682 | &EXT4_SB(sb)->s_freeclusters_counter))); | |
d5e03cbb TT |
4683 | if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter)) |
4684 | es->s_free_inodes_count = | |
4685 | cpu_to_le32(percpu_counter_sum_positive( | |
ce7e010a | 4686 | &EXT4_SB(sb)->s_freeinodes_counter)); |
ac27a0ec | 4687 | BUFFER_TRACE(sbh, "marking dirty"); |
06db49e6 | 4688 | ext4_superblock_csum_set(sb); |
1566a48a TT |
4689 | if (sync) |
4690 | lock_buffer(sbh); | |
4743f839 PKS |
4691 | if (buffer_write_io_error(sbh)) { |
4692 | /* | |
4693 | * Oh, dear. A previous attempt to write the | |
4694 | * superblock failed. This could happen because the | |
4695 | * USB device was yanked out. Or it could happen to | |
4696 | * be a transient write error and maybe the block will | |
4697 | * be remapped. Nothing we can do but to retry the | |
4698 | * write and hope for the best. | |
4699 | */ | |
4700 | ext4_msg(sb, KERN_ERR, "previous I/O error to " | |
4701 | "superblock detected"); | |
4702 | clear_buffer_write_io_error(sbh); | |
4703 | set_buffer_uptodate(sbh); | |
4704 | } | |
ac27a0ec | 4705 | mark_buffer_dirty(sbh); |
914258bf | 4706 | if (sync) { |
1566a48a | 4707 | unlock_buffer(sbh); |
564bc402 | 4708 | error = __sync_dirty_buffer(sbh, |
00473374 | 4709 | REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0)); |
c4be0c1d TS |
4710 | if (error) |
4711 | return error; | |
4712 | ||
4713 | error = buffer_write_io_error(sbh); | |
4714 | if (error) { | |
b31e1552 ES |
4715 | ext4_msg(sb, KERN_ERR, "I/O error while writing " |
4716 | "superblock"); | |
914258bf TT |
4717 | clear_buffer_write_io_error(sbh); |
4718 | set_buffer_uptodate(sbh); | |
4719 | } | |
4720 | } | |
c4be0c1d | 4721 | return error; |
ac27a0ec DK |
4722 | } |
4723 | ||
ac27a0ec DK |
4724 | /* |
4725 | * Have we just finished recovery? If so, and if we are mounting (or | |
4726 | * remounting) the filesystem readonly, then we will end up with a | |
4727 | * consistent fs on disk. Record that fact. | |
4728 | */ | |
2b2d6d01 TT |
4729 | static void ext4_mark_recovery_complete(struct super_block *sb, |
4730 | struct ext4_super_block *es) | |
ac27a0ec | 4731 | { |
617ba13b | 4732 | journal_t *journal = EXT4_SB(sb)->s_journal; |
ac27a0ec | 4733 | |
e2b911c5 | 4734 | if (!ext4_has_feature_journal(sb)) { |
0390131b FM |
4735 | BUG_ON(journal != NULL); |
4736 | return; | |
4737 | } | |
dab291af | 4738 | jbd2_journal_lock_updates(journal); |
7ffe1ea8 HK |
4739 | if (jbd2_journal_flush(journal) < 0) |
4740 | goto out; | |
4741 | ||
bc98a42c | 4742 | if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) { |
e2b911c5 | 4743 | ext4_clear_feature_journal_needs_recovery(sb); |
e2d67052 | 4744 | ext4_commit_super(sb, 1); |
ac27a0ec | 4745 | } |
7ffe1ea8 HK |
4746 | |
4747 | out: | |
dab291af | 4748 | jbd2_journal_unlock_updates(journal); |
ac27a0ec DK |
4749 | } |
4750 | ||
4751 | /* | |
4752 | * If we are mounting (or read-write remounting) a filesystem whose journal | |
4753 | * has recorded an error from a previous lifetime, move that error to the | |
4754 | * main filesystem now. | |
4755 | */ | |
2b2d6d01 TT |
4756 | static void ext4_clear_journal_err(struct super_block *sb, |
4757 | struct ext4_super_block *es) | |
ac27a0ec DK |
4758 | { |
4759 | journal_t *journal; | |
4760 | int j_errno; | |
4761 | const char *errstr; | |
4762 | ||
e2b911c5 | 4763 | BUG_ON(!ext4_has_feature_journal(sb)); |
0390131b | 4764 | |
617ba13b | 4765 | journal = EXT4_SB(sb)->s_journal; |
ac27a0ec DK |
4766 | |
4767 | /* | |
4768 | * Now check for any error status which may have been recorded in the | |
617ba13b | 4769 | * journal by a prior ext4_error() or ext4_abort() |
ac27a0ec DK |
4770 | */ |
4771 | ||
dab291af | 4772 | j_errno = jbd2_journal_errno(journal); |
ac27a0ec DK |
4773 | if (j_errno) { |
4774 | char nbuf[16]; | |
4775 | ||
617ba13b | 4776 | errstr = ext4_decode_error(sb, j_errno, nbuf); |
12062ddd | 4777 | ext4_warning(sb, "Filesystem error recorded " |
ac27a0ec | 4778 | "from previous mount: %s", errstr); |
12062ddd | 4779 | ext4_warning(sb, "Marking fs in need of filesystem check."); |
ac27a0ec | 4780 | |
617ba13b MC |
4781 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; |
4782 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
e2d67052 | 4783 | ext4_commit_super(sb, 1); |
ac27a0ec | 4784 | |
dab291af | 4785 | jbd2_journal_clear_err(journal); |
d796c52e | 4786 | jbd2_journal_update_sb_errno(journal); |
ac27a0ec DK |
4787 | } |
4788 | } | |
4789 | ||
4790 | /* | |
4791 | * Force the running and committing transactions to commit, | |
4792 | * and wait on the commit. | |
4793 | */ | |
617ba13b | 4794 | int ext4_force_commit(struct super_block *sb) |
ac27a0ec DK |
4795 | { |
4796 | journal_t *journal; | |
ac27a0ec | 4797 | |
bc98a42c | 4798 | if (sb_rdonly(sb)) |
ac27a0ec DK |
4799 | return 0; |
4800 | ||
617ba13b | 4801 | journal = EXT4_SB(sb)->s_journal; |
b1deefc9 | 4802 | return ext4_journal_force_commit(journal); |
ac27a0ec DK |
4803 | } |
4804 | ||
617ba13b | 4805 | static int ext4_sync_fs(struct super_block *sb, int wait) |
ac27a0ec | 4806 | { |
14ce0cb4 | 4807 | int ret = 0; |
9eddacf9 | 4808 | tid_t target; |
06a407f1 | 4809 | bool needs_barrier = false; |
8d5d02e6 | 4810 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 4811 | |
0db1ff22 TT |
4812 | if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) |
4813 | return 0; | |
4814 | ||
9bffad1e | 4815 | trace_ext4_sync_fs(sb, wait); |
2e8fa54e | 4816 | flush_workqueue(sbi->rsv_conversion_wq); |
a1177825 JK |
4817 | /* |
4818 | * Writeback quota in non-journalled quota case - journalled quota has | |
4819 | * no dirty dquots | |
4820 | */ | |
4821 | dquot_writeback_dquots(sb, -1); | |
06a407f1 DM |
4822 | /* |
4823 | * Data writeback is possible w/o journal transaction, so barrier must | |
4824 | * being sent at the end of the function. But we can skip it if | |
4825 | * transaction_commit will do it for us. | |
4826 | */ | |
bda32530 TT |
4827 | if (sbi->s_journal) { |
4828 | target = jbd2_get_latest_transaction(sbi->s_journal); | |
4829 | if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && | |
4830 | !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) | |
4831 | needs_barrier = true; | |
4832 | ||
4833 | if (jbd2_journal_start_commit(sbi->s_journal, &target)) { | |
4834 | if (wait) | |
4835 | ret = jbd2_log_wait_commit(sbi->s_journal, | |
4836 | target); | |
4837 | } | |
4838 | } else if (wait && test_opt(sb, BARRIER)) | |
06a407f1 | 4839 | needs_barrier = true; |
06a407f1 DM |
4840 | if (needs_barrier) { |
4841 | int err; | |
4842 | err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL); | |
4843 | if (!ret) | |
4844 | ret = err; | |
0390131b | 4845 | } |
06a407f1 DM |
4846 | |
4847 | return ret; | |
4848 | } | |
4849 | ||
ac27a0ec DK |
4850 | /* |
4851 | * LVM calls this function before a (read-only) snapshot is created. This | |
4852 | * gives us a chance to flush the journal completely and mark the fs clean. | |
be4f27d3 YY |
4853 | * |
4854 | * Note that only this function cannot bring a filesystem to be in a clean | |
8e8ad8a5 JK |
4855 | * state independently. It relies on upper layer to stop all data & metadata |
4856 | * modifications. | |
ac27a0ec | 4857 | */ |
c4be0c1d | 4858 | static int ext4_freeze(struct super_block *sb) |
ac27a0ec | 4859 | { |
c4be0c1d TS |
4860 | int error = 0; |
4861 | journal_t *journal; | |
ac27a0ec | 4862 | |
bc98a42c | 4863 | if (sb_rdonly(sb)) |
9ca92389 | 4864 | return 0; |
ac27a0ec | 4865 | |
9ca92389 | 4866 | journal = EXT4_SB(sb)->s_journal; |
7ffe1ea8 | 4867 | |
bb044576 TT |
4868 | if (journal) { |
4869 | /* Now we set up the journal barrier. */ | |
4870 | jbd2_journal_lock_updates(journal); | |
ac27a0ec | 4871 | |
bb044576 TT |
4872 | /* |
4873 | * Don't clear the needs_recovery flag if we failed to | |
4874 | * flush the journal. | |
4875 | */ | |
4876 | error = jbd2_journal_flush(journal); | |
4877 | if (error < 0) | |
4878 | goto out; | |
c642dc9e ES |
4879 | |
4880 | /* Journal blocked and flushed, clear needs_recovery flag. */ | |
e2b911c5 | 4881 | ext4_clear_feature_journal_needs_recovery(sb); |
bb044576 | 4882 | } |
9ca92389 | 4883 | |
9ca92389 | 4884 | error = ext4_commit_super(sb, 1); |
6b0310fb | 4885 | out: |
bb044576 TT |
4886 | if (journal) |
4887 | /* we rely on upper layer to stop further updates */ | |
4888 | jbd2_journal_unlock_updates(journal); | |
6b0310fb | 4889 | return error; |
ac27a0ec DK |
4890 | } |
4891 | ||
4892 | /* | |
4893 | * Called by LVM after the snapshot is done. We need to reset the RECOVER | |
4894 | * flag here, even though the filesystem is not technically dirty yet. | |
4895 | */ | |
c4be0c1d | 4896 | static int ext4_unfreeze(struct super_block *sb) |
ac27a0ec | 4897 | { |
bc98a42c | 4898 | if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb))) |
9ca92389 TT |
4899 | return 0; |
4900 | ||
c642dc9e ES |
4901 | if (EXT4_SB(sb)->s_journal) { |
4902 | /* Reset the needs_recovery flag before the fs is unlocked. */ | |
e2b911c5 | 4903 | ext4_set_feature_journal_needs_recovery(sb); |
c642dc9e ES |
4904 | } |
4905 | ||
9ca92389 | 4906 | ext4_commit_super(sb, 1); |
c4be0c1d | 4907 | return 0; |
ac27a0ec DK |
4908 | } |
4909 | ||
673c6100 TT |
4910 | /* |
4911 | * Structure to save mount options for ext4_remount's benefit | |
4912 | */ | |
4913 | struct ext4_mount_options { | |
4914 | unsigned long s_mount_opt; | |
a2595b8a | 4915 | unsigned long s_mount_opt2; |
08cefc7a EB |
4916 | kuid_t s_resuid; |
4917 | kgid_t s_resgid; | |
673c6100 TT |
4918 | unsigned long s_commit_interval; |
4919 | u32 s_min_batch_time, s_max_batch_time; | |
4920 | #ifdef CONFIG_QUOTA | |
4921 | int s_jquota_fmt; | |
a2d4a646 | 4922 | char *s_qf_names[EXT4_MAXQUOTAS]; |
673c6100 TT |
4923 | #endif |
4924 | }; | |
4925 | ||
2b2d6d01 | 4926 | static int ext4_remount(struct super_block *sb, int *flags, char *data) |
ac27a0ec | 4927 | { |
2b2d6d01 | 4928 | struct ext4_super_block *es; |
617ba13b | 4929 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 4930 | unsigned long old_sb_flags; |
617ba13b | 4931 | struct ext4_mount_options old_opts; |
c79d967d | 4932 | int enable_quota = 0; |
8a266467 | 4933 | ext4_group_t g; |
b3881f74 | 4934 | unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; |
c5e06d10 | 4935 | int err = 0; |
ac27a0ec | 4936 | #ifdef CONFIG_QUOTA |
03dafb5f | 4937 | int i, j; |
ac27a0ec | 4938 | #endif |
d4c402d9 | 4939 | char *orig_data = kstrdup(data, GFP_KERNEL); |
ac27a0ec DK |
4940 | |
4941 | /* Store the original options */ | |
4942 | old_sb_flags = sb->s_flags; | |
4943 | old_opts.s_mount_opt = sbi->s_mount_opt; | |
a2595b8a | 4944 | old_opts.s_mount_opt2 = sbi->s_mount_opt2; |
ac27a0ec DK |
4945 | old_opts.s_resuid = sbi->s_resuid; |
4946 | old_opts.s_resgid = sbi->s_resgid; | |
4947 | old_opts.s_commit_interval = sbi->s_commit_interval; | |
30773840 TT |
4948 | old_opts.s_min_batch_time = sbi->s_min_batch_time; |
4949 | old_opts.s_max_batch_time = sbi->s_max_batch_time; | |
ac27a0ec DK |
4950 | #ifdef CONFIG_QUOTA |
4951 | old_opts.s_jquota_fmt = sbi->s_jquota_fmt; | |
a2d4a646 | 4952 | for (i = 0; i < EXT4_MAXQUOTAS; i++) |
03dafb5f CG |
4953 | if (sbi->s_qf_names[i]) { |
4954 | old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i], | |
4955 | GFP_KERNEL); | |
4956 | if (!old_opts.s_qf_names[i]) { | |
4957 | for (j = 0; j < i; j++) | |
4958 | kfree(old_opts.s_qf_names[j]); | |
3e36a163 | 4959 | kfree(orig_data); |
03dafb5f CG |
4960 | return -ENOMEM; |
4961 | } | |
4962 | } else | |
4963 | old_opts.s_qf_names[i] = NULL; | |
ac27a0ec | 4964 | #endif |
b3881f74 TT |
4965 | if (sbi->s_journal && sbi->s_journal->j_task->io_context) |
4966 | journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; | |
ac27a0ec | 4967 | |
661aa520 | 4968 | if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) { |
ac27a0ec DK |
4969 | err = -EINVAL; |
4970 | goto restore_opts; | |
4971 | } | |
4972 | ||
6b992ff2 | 4973 | if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ |
c6d3d56d DW |
4974 | test_opt(sb, JOURNAL_CHECKSUM)) { |
4975 | ext4_msg(sb, KERN_ERR, "changing journal_checksum " | |
2d5b86e0 ES |
4976 | "during remount not supported; ignoring"); |
4977 | sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; | |
6b992ff2 DW |
4978 | } |
4979 | ||
6ae6514b PS |
4980 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { |
4981 | if (test_opt2(sb, EXPLICIT_DELALLOC)) { | |
4982 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
4983 | "both data=journal and delalloc"); | |
4984 | err = -EINVAL; | |
4985 | goto restore_opts; | |
4986 | } | |
4987 | if (test_opt(sb, DIOREAD_NOLOCK)) { | |
4988 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
4989 | "both data=journal and dioread_nolock"); | |
4990 | err = -EINVAL; | |
4991 | goto restore_opts; | |
4992 | } | |
923ae0ff RZ |
4993 | if (test_opt(sb, DAX)) { |
4994 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
4995 | "both data=journal and dax"); | |
4996 | err = -EINVAL; | |
4997 | goto restore_opts; | |
4998 | } | |
ab04df78 JK |
4999 | } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) { |
5000 | if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { | |
5001 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
5002 | "journal_async_commit in data=ordered mode"); | |
5003 | err = -EINVAL; | |
5004 | goto restore_opts; | |
5005 | } | |
923ae0ff RZ |
5006 | } |
5007 | ||
cdb7ee4c TE |
5008 | if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) { |
5009 | ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount"); | |
5010 | err = -EINVAL; | |
5011 | goto restore_opts; | |
5012 | } | |
5013 | ||
923ae0ff RZ |
5014 | if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) { |
5015 | ext4_msg(sb, KERN_WARNING, "warning: refusing change of " | |
5016 | "dax flag with busy inodes while remounting"); | |
5017 | sbi->s_mount_opt ^= EXT4_MOUNT_DAX; | |
6ae6514b PS |
5018 | } |
5019 | ||
4ab2f15b | 5020 | if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) |
c67d859e | 5021 | ext4_abort(sb, "Abort forced by user"); |
ac27a0ec DK |
5022 | |
5023 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
482a7425 | 5024 | (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); |
ac27a0ec DK |
5025 | |
5026 | es = sbi->s_es; | |
5027 | ||
b3881f74 | 5028 | if (sbi->s_journal) { |
0390131b | 5029 | ext4_init_journal_params(sb, sbi->s_journal); |
b3881f74 TT |
5030 | set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); |
5031 | } | |
ac27a0ec | 5032 | |
a2fd66d0 TT |
5033 | if (*flags & MS_LAZYTIME) |
5034 | sb->s_flags |= MS_LAZYTIME; | |
5035 | ||
bc98a42c | 5036 | if ((bool)(*flags & MS_RDONLY) != sb_rdonly(sb)) { |
4ab2f15b | 5037 | if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { |
ac27a0ec DK |
5038 | err = -EROFS; |
5039 | goto restore_opts; | |
5040 | } | |
5041 | ||
5042 | if (*flags & MS_RDONLY) { | |
38c03b34 TT |
5043 | err = sync_filesystem(sb); |
5044 | if (err < 0) | |
5045 | goto restore_opts; | |
0f0dd62f CH |
5046 | err = dquot_suspend(sb, -1); |
5047 | if (err < 0) | |
c79d967d | 5048 | goto restore_opts; |
c79d967d | 5049 | |
ac27a0ec DK |
5050 | /* |
5051 | * First of all, the unconditional stuff we have to do | |
5052 | * to disable replay of the journal when we next remount | |
5053 | */ | |
5054 | sb->s_flags |= MS_RDONLY; | |
5055 | ||
5056 | /* | |
5057 | * OK, test if we are remounting a valid rw partition | |
5058 | * readonly, and if so set the rdonly flag and then | |
5059 | * mark the partition as valid again. | |
5060 | */ | |
617ba13b MC |
5061 | if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && |
5062 | (sbi->s_mount_state & EXT4_VALID_FS)) | |
ac27a0ec DK |
5063 | es->s_state = cpu_to_le16(sbi->s_mount_state); |
5064 | ||
a63c9eb2 | 5065 | if (sbi->s_journal) |
0390131b | 5066 | ext4_mark_recovery_complete(sb, es); |
ac27a0ec | 5067 | } else { |
a13fb1a4 | 5068 | /* Make sure we can mount this feature set readwrite */ |
e2b911c5 | 5069 | if (ext4_has_feature_readonly(sb) || |
2cb5cc8b | 5070 | !ext4_feature_set_ok(sb, 0)) { |
ac27a0ec DK |
5071 | err = -EROFS; |
5072 | goto restore_opts; | |
5073 | } | |
8a266467 TT |
5074 | /* |
5075 | * Make sure the group descriptor checksums | |
0b8e58a1 | 5076 | * are sane. If they aren't, refuse to remount r/w. |
8a266467 TT |
5077 | */ |
5078 | for (g = 0; g < sbi->s_groups_count; g++) { | |
5079 | struct ext4_group_desc *gdp = | |
5080 | ext4_get_group_desc(sb, g, NULL); | |
5081 | ||
feb0ab32 | 5082 | if (!ext4_group_desc_csum_verify(sb, g, gdp)) { |
b31e1552 ES |
5083 | ext4_msg(sb, KERN_ERR, |
5084 | "ext4_remount: Checksum for group %u failed (%u!=%u)", | |
e2b911c5 | 5085 | g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), |
8a266467 | 5086 | le16_to_cpu(gdp->bg_checksum)); |
6a797d27 | 5087 | err = -EFSBADCRC; |
8a266467 TT |
5088 | goto restore_opts; |
5089 | } | |
5090 | } | |
5091 | ||
ead6596b ES |
5092 | /* |
5093 | * If we have an unprocessed orphan list hanging | |
5094 | * around from a previously readonly bdev mount, | |
5095 | * require a full umount/remount for now. | |
5096 | */ | |
5097 | if (es->s_last_orphan) { | |
b31e1552 | 5098 | ext4_msg(sb, KERN_WARNING, "Couldn't " |
ead6596b ES |
5099 | "remount RDWR because of unprocessed " |
5100 | "orphan inode list. Please " | |
b31e1552 | 5101 | "umount/remount instead"); |
ead6596b ES |
5102 | err = -EINVAL; |
5103 | goto restore_opts; | |
5104 | } | |
5105 | ||
ac27a0ec DK |
5106 | /* |
5107 | * Mounting a RDONLY partition read-write, so reread | |
5108 | * and store the current valid flag. (It may have | |
5109 | * been changed by e2fsck since we originally mounted | |
5110 | * the partition.) | |
5111 | */ | |
0390131b FM |
5112 | if (sbi->s_journal) |
5113 | ext4_clear_journal_err(sb, es); | |
ac27a0ec | 5114 | sbi->s_mount_state = le16_to_cpu(es->s_state); |
2b2d6d01 | 5115 | if (!ext4_setup_super(sb, es, 0)) |
ac27a0ec | 5116 | sb->s_flags &= ~MS_RDONLY; |
e2b911c5 | 5117 | if (ext4_has_feature_mmp(sb)) |
c5e06d10 JL |
5118 | if (ext4_multi_mount_protect(sb, |
5119 | le64_to_cpu(es->s_mmp_block))) { | |
5120 | err = -EROFS; | |
5121 | goto restore_opts; | |
5122 | } | |
c79d967d | 5123 | enable_quota = 1; |
ac27a0ec DK |
5124 | } |
5125 | } | |
bfff6873 LC |
5126 | |
5127 | /* | |
5128 | * Reinitialize lazy itable initialization thread based on | |
5129 | * current settings | |
5130 | */ | |
bc98a42c | 5131 | if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE)) |
bfff6873 LC |
5132 | ext4_unregister_li_request(sb); |
5133 | else { | |
5134 | ext4_group_t first_not_zeroed; | |
5135 | first_not_zeroed = ext4_has_uninit_itable(sb); | |
5136 | ext4_register_li_request(sb, first_not_zeroed); | |
5137 | } | |
5138 | ||
6fd058f7 | 5139 | ext4_setup_system_zone(sb); |
d096ad0f | 5140 | if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY)) |
e2d67052 | 5141 | ext4_commit_super(sb, 1); |
0390131b | 5142 | |
ac27a0ec DK |
5143 | #ifdef CONFIG_QUOTA |
5144 | /* Release old quota file names */ | |
a2d4a646 | 5145 | for (i = 0; i < EXT4_MAXQUOTAS; i++) |
03dafb5f | 5146 | kfree(old_opts.s_qf_names[i]); |
7c319d32 AK |
5147 | if (enable_quota) { |
5148 | if (sb_any_quota_suspended(sb)) | |
5149 | dquot_resume(sb, -1); | |
e2b911c5 | 5150 | else if (ext4_has_feature_quota(sb)) { |
7c319d32 | 5151 | err = ext4_enable_quotas(sb); |
07724f98 | 5152 | if (err) |
7c319d32 | 5153 | goto restore_opts; |
7c319d32 AK |
5154 | } |
5155 | } | |
ac27a0ec | 5156 | #endif |
d4c402d9 | 5157 | |
a26f4992 | 5158 | *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME); |
d4c402d9 CW |
5159 | ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); |
5160 | kfree(orig_data); | |
ac27a0ec | 5161 | return 0; |
0b8e58a1 | 5162 | |
ac27a0ec DK |
5163 | restore_opts: |
5164 | sb->s_flags = old_sb_flags; | |
5165 | sbi->s_mount_opt = old_opts.s_mount_opt; | |
a2595b8a | 5166 | sbi->s_mount_opt2 = old_opts.s_mount_opt2; |
ac27a0ec DK |
5167 | sbi->s_resuid = old_opts.s_resuid; |
5168 | sbi->s_resgid = old_opts.s_resgid; | |
5169 | sbi->s_commit_interval = old_opts.s_commit_interval; | |
30773840 TT |
5170 | sbi->s_min_batch_time = old_opts.s_min_batch_time; |
5171 | sbi->s_max_batch_time = old_opts.s_max_batch_time; | |
ac27a0ec DK |
5172 | #ifdef CONFIG_QUOTA |
5173 | sbi->s_jquota_fmt = old_opts.s_jquota_fmt; | |
a2d4a646 | 5174 | for (i = 0; i < EXT4_MAXQUOTAS; i++) { |
03dafb5f | 5175 | kfree(sbi->s_qf_names[i]); |
ac27a0ec DK |
5176 | sbi->s_qf_names[i] = old_opts.s_qf_names[i]; |
5177 | } | |
5178 | #endif | |
d4c402d9 | 5179 | kfree(orig_data); |
ac27a0ec DK |
5180 | return err; |
5181 | } | |
5182 | ||
689c958c LX |
5183 | #ifdef CONFIG_QUOTA |
5184 | static int ext4_statfs_project(struct super_block *sb, | |
5185 | kprojid_t projid, struct kstatfs *buf) | |
5186 | { | |
5187 | struct kqid qid; | |
5188 | struct dquot *dquot; | |
5189 | u64 limit; | |
5190 | u64 curblock; | |
5191 | ||
5192 | qid = make_kqid_projid(projid); | |
5193 | dquot = dqget(sb, qid); | |
5194 | if (IS_ERR(dquot)) | |
5195 | return PTR_ERR(dquot); | |
5196 | spin_lock(&dq_data_lock); | |
5197 | ||
5198 | limit = (dquot->dq_dqb.dqb_bsoftlimit ? | |
5199 | dquot->dq_dqb.dqb_bsoftlimit : | |
5200 | dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits; | |
5201 | if (limit && buf->f_blocks > limit) { | |
5202 | curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits; | |
5203 | buf->f_blocks = limit; | |
5204 | buf->f_bfree = buf->f_bavail = | |
5205 | (buf->f_blocks > curblock) ? | |
5206 | (buf->f_blocks - curblock) : 0; | |
5207 | } | |
5208 | ||
5209 | limit = dquot->dq_dqb.dqb_isoftlimit ? | |
5210 | dquot->dq_dqb.dqb_isoftlimit : | |
5211 | dquot->dq_dqb.dqb_ihardlimit; | |
5212 | if (limit && buf->f_files > limit) { | |
5213 | buf->f_files = limit; | |
5214 | buf->f_ffree = | |
5215 | (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? | |
5216 | (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; | |
5217 | } | |
5218 | ||
5219 | spin_unlock(&dq_data_lock); | |
5220 | dqput(dquot); | |
5221 | return 0; | |
5222 | } | |
5223 | #endif | |
5224 | ||
2b2d6d01 | 5225 | static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) |
ac27a0ec DK |
5226 | { |
5227 | struct super_block *sb = dentry->d_sb; | |
617ba13b MC |
5228 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
5229 | struct ext4_super_block *es = sbi->s_es; | |
27dd4385 | 5230 | ext4_fsblk_t overhead = 0, resv_blocks; |
960cc398 | 5231 | u64 fsid; |
d02a9391 | 5232 | s64 bfree; |
27dd4385 | 5233 | resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); |
ac27a0ec | 5234 | |
952fc18e TT |
5235 | if (!test_opt(sb, MINIX_DF)) |
5236 | overhead = sbi->s_overhead; | |
ac27a0ec | 5237 | |
617ba13b | 5238 | buf->f_type = EXT4_SUPER_MAGIC; |
ac27a0ec | 5239 | buf->f_bsize = sb->s_blocksize; |
b72f78cb | 5240 | buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); |
57042651 TT |
5241 | bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - |
5242 | percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); | |
d02a9391 | 5243 | /* prevent underflow in case that few free space is available */ |
57042651 | 5244 | buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); |
27dd4385 LC |
5245 | buf->f_bavail = buf->f_bfree - |
5246 | (ext4_r_blocks_count(es) + resv_blocks); | |
5247 | if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) | |
ac27a0ec DK |
5248 | buf->f_bavail = 0; |
5249 | buf->f_files = le32_to_cpu(es->s_inodes_count); | |
52d9f3b4 | 5250 | buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); |
617ba13b | 5251 | buf->f_namelen = EXT4_NAME_LEN; |
960cc398 PE |
5252 | fsid = le64_to_cpup((void *)es->s_uuid) ^ |
5253 | le64_to_cpup((void *)es->s_uuid + sizeof(u64)); | |
5254 | buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; | |
5255 | buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; | |
0b8e58a1 | 5256 | |
689c958c LX |
5257 | #ifdef CONFIG_QUOTA |
5258 | if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && | |
5259 | sb_has_quota_limits_enabled(sb, PRJQUOTA)) | |
5260 | ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); | |
5261 | #endif | |
ac27a0ec DK |
5262 | return 0; |
5263 | } | |
5264 | ||
0b8e58a1 AD |
5265 | /* Helper function for writing quotas on sync - we need to start transaction |
5266 | * before quota file is locked for write. Otherwise the are possible deadlocks: | |
ac27a0ec | 5267 | * Process 1 Process 2 |
617ba13b | 5268 | * ext4_create() quota_sync() |
a269eb18 | 5269 | * jbd2_journal_start() write_dquot() |
871a2931 | 5270 | * dquot_initialize() down(dqio_mutex) |
dab291af | 5271 | * down(dqio_mutex) jbd2_journal_start() |
ac27a0ec DK |
5272 | * |
5273 | */ | |
5274 | ||
5275 | #ifdef CONFIG_QUOTA | |
5276 | ||
5277 | static inline struct inode *dquot_to_inode(struct dquot *dquot) | |
5278 | { | |
4c376dca | 5279 | return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; |
ac27a0ec DK |
5280 | } |
5281 | ||
617ba13b | 5282 | static int ext4_write_dquot(struct dquot *dquot) |
ac27a0ec DK |
5283 | { |
5284 | int ret, err; | |
5285 | handle_t *handle; | |
5286 | struct inode *inode; | |
5287 | ||
5288 | inode = dquot_to_inode(dquot); | |
9924a92a | 5289 | handle = ext4_journal_start(inode, EXT4_HT_QUOTA, |
0b8e58a1 | 5290 | EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); |
ac27a0ec DK |
5291 | if (IS_ERR(handle)) |
5292 | return PTR_ERR(handle); | |
5293 | ret = dquot_commit(dquot); | |
617ba13b | 5294 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
5295 | if (!ret) |
5296 | ret = err; | |
5297 | return ret; | |
5298 | } | |
5299 | ||
617ba13b | 5300 | static int ext4_acquire_dquot(struct dquot *dquot) |
ac27a0ec DK |
5301 | { |
5302 | int ret, err; | |
5303 | handle_t *handle; | |
5304 | ||
9924a92a | 5305 | handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, |
0b8e58a1 | 5306 | EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); |
ac27a0ec DK |
5307 | if (IS_ERR(handle)) |
5308 | return PTR_ERR(handle); | |
5309 | ret = dquot_acquire(dquot); | |
617ba13b | 5310 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
5311 | if (!ret) |
5312 | ret = err; | |
5313 | return ret; | |
5314 | } | |
5315 | ||
617ba13b | 5316 | static int ext4_release_dquot(struct dquot *dquot) |
ac27a0ec DK |
5317 | { |
5318 | int ret, err; | |
5319 | handle_t *handle; | |
5320 | ||
9924a92a | 5321 | handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, |
0b8e58a1 | 5322 | EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); |
9c3013e9 JK |
5323 | if (IS_ERR(handle)) { |
5324 | /* Release dquot anyway to avoid endless cycle in dqput() */ | |
5325 | dquot_release(dquot); | |
ac27a0ec | 5326 | return PTR_ERR(handle); |
9c3013e9 | 5327 | } |
ac27a0ec | 5328 | ret = dquot_release(dquot); |
617ba13b | 5329 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
5330 | if (!ret) |
5331 | ret = err; | |
5332 | return ret; | |
5333 | } | |
5334 | ||
617ba13b | 5335 | static int ext4_mark_dquot_dirty(struct dquot *dquot) |
ac27a0ec | 5336 | { |
262b4662 JK |
5337 | struct super_block *sb = dquot->dq_sb; |
5338 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
5339 | ||
2c8be6b2 | 5340 | /* Are we journaling quotas? */ |
e2b911c5 | 5341 | if (ext4_has_feature_quota(sb) || |
262b4662 | 5342 | sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { |
ac27a0ec | 5343 | dquot_mark_dquot_dirty(dquot); |
617ba13b | 5344 | return ext4_write_dquot(dquot); |
ac27a0ec DK |
5345 | } else { |
5346 | return dquot_mark_dquot_dirty(dquot); | |
5347 | } | |
5348 | } | |
5349 | ||
617ba13b | 5350 | static int ext4_write_info(struct super_block *sb, int type) |
ac27a0ec DK |
5351 | { |
5352 | int ret, err; | |
5353 | handle_t *handle; | |
5354 | ||
5355 | /* Data block + inode block */ | |
2b0143b5 | 5356 | handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2); |
ac27a0ec DK |
5357 | if (IS_ERR(handle)) |
5358 | return PTR_ERR(handle); | |
5359 | ret = dquot_commit_info(sb, type); | |
617ba13b | 5360 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
5361 | if (!ret) |
5362 | ret = err; | |
5363 | return ret; | |
5364 | } | |
5365 | ||
5366 | /* | |
5367 | * Turn on quotas during mount time - we need to find | |
5368 | * the quota file and such... | |
5369 | */ | |
617ba13b | 5370 | static int ext4_quota_on_mount(struct super_block *sb, int type) |
ac27a0ec | 5371 | { |
287a8095 CH |
5372 | return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type], |
5373 | EXT4_SB(sb)->s_jquota_fmt, type); | |
ac27a0ec DK |
5374 | } |
5375 | ||
daf647d2 TT |
5376 | static void lockdep_set_quota_inode(struct inode *inode, int subclass) |
5377 | { | |
5378 | struct ext4_inode_info *ei = EXT4_I(inode); | |
5379 | ||
5380 | /* The first argument of lockdep_set_subclass has to be | |
5381 | * *exactly* the same as the argument to init_rwsem() --- in | |
5382 | * this case, in init_once() --- or lockdep gets unhappy | |
5383 | * because the name of the lock is set using the | |
5384 | * stringification of the argument to init_rwsem(). | |
5385 | */ | |
5386 | (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ | |
5387 | lockdep_set_subclass(&ei->i_data_sem, subclass); | |
5388 | } | |
5389 | ||
ac27a0ec DK |
5390 | /* |
5391 | * Standard function to be called on quota_on | |
5392 | */ | |
617ba13b | 5393 | static int ext4_quota_on(struct super_block *sb, int type, int format_id, |
8c54ca9c | 5394 | const struct path *path) |
ac27a0ec DK |
5395 | { |
5396 | int err; | |
ac27a0ec DK |
5397 | |
5398 | if (!test_opt(sb, QUOTA)) | |
5399 | return -EINVAL; | |
0623543b | 5400 | |
ac27a0ec | 5401 | /* Quotafile not on the same filesystem? */ |
d8c9584e | 5402 | if (path->dentry->d_sb != sb) |
ac27a0ec | 5403 | return -EXDEV; |
0623543b JK |
5404 | /* Journaling quota? */ |
5405 | if (EXT4_SB(sb)->s_qf_names[type]) { | |
2b2d6d01 | 5406 | /* Quotafile not in fs root? */ |
f00c9e44 | 5407 | if (path->dentry->d_parent != sb->s_root) |
b31e1552 ES |
5408 | ext4_msg(sb, KERN_WARNING, |
5409 | "Quota file not on filesystem root. " | |
5410 | "Journaled quota will not work"); | |
2b2d6d01 | 5411 | } |
0623543b JK |
5412 | |
5413 | /* | |
5414 | * When we journal data on quota file, we have to flush journal to see | |
5415 | * all updates to the file when we bypass pagecache... | |
5416 | */ | |
0390131b | 5417 | if (EXT4_SB(sb)->s_journal && |
2b0143b5 | 5418 | ext4_should_journal_data(d_inode(path->dentry))) { |
0623543b JK |
5419 | /* |
5420 | * We don't need to lock updates but journal_flush() could | |
5421 | * otherwise be livelocked... | |
5422 | */ | |
5423 | jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); | |
7ffe1ea8 | 5424 | err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); |
0623543b | 5425 | jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); |
f00c9e44 | 5426 | if (err) |
7ffe1ea8 | 5427 | return err; |
0623543b | 5428 | } |
957153fc | 5429 | |
daf647d2 TT |
5430 | lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); |
5431 | err = dquot_quota_on(sb, type, format_id, path); | |
957153fc | 5432 | if (err) { |
daf647d2 TT |
5433 | lockdep_set_quota_inode(path->dentry->d_inode, |
5434 | I_DATA_SEM_NORMAL); | |
957153fc JK |
5435 | } else { |
5436 | struct inode *inode = d_inode(path->dentry); | |
5437 | handle_t *handle; | |
5438 | ||
61a92987 JK |
5439 | /* |
5440 | * Set inode flags to prevent userspace from messing with quota | |
5441 | * files. If this fails, we return success anyway since quotas | |
5442 | * are already enabled and this is not a hard failure. | |
5443 | */ | |
957153fc JK |
5444 | inode_lock(inode); |
5445 | handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); | |
5446 | if (IS_ERR(handle)) | |
5447 | goto unlock_inode; | |
5448 | EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL; | |
5449 | inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, | |
5450 | S_NOATIME | S_IMMUTABLE); | |
5451 | ext4_mark_inode_dirty(handle, inode); | |
5452 | ext4_journal_stop(handle); | |
5453 | unlock_inode: | |
5454 | inode_unlock(inode); | |
5455 | } | |
daf647d2 | 5456 | return err; |
ac27a0ec DK |
5457 | } |
5458 | ||
7c319d32 AK |
5459 | static int ext4_quota_enable(struct super_block *sb, int type, int format_id, |
5460 | unsigned int flags) | |
5461 | { | |
5462 | int err; | |
5463 | struct inode *qf_inode; | |
a2d4a646 | 5464 | unsigned long qf_inums[EXT4_MAXQUOTAS] = { |
7c319d32 | 5465 | le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), |
689c958c LX |
5466 | le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), |
5467 | le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) | |
7c319d32 AK |
5468 | }; |
5469 | ||
e2b911c5 | 5470 | BUG_ON(!ext4_has_feature_quota(sb)); |
7c319d32 AK |
5471 | |
5472 | if (!qf_inums[type]) | |
5473 | return -EPERM; | |
5474 | ||
5475 | qf_inode = ext4_iget(sb, qf_inums[type]); | |
5476 | if (IS_ERR(qf_inode)) { | |
5477 | ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); | |
5478 | return PTR_ERR(qf_inode); | |
5479 | } | |
5480 | ||
bcb13850 JK |
5481 | /* Don't account quota for quota files to avoid recursion */ |
5482 | qf_inode->i_flags |= S_NOQUOTA; | |
daf647d2 | 5483 | lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); |
7c319d32 AK |
5484 | err = dquot_enable(qf_inode, type, format_id, flags); |
5485 | iput(qf_inode); | |
daf647d2 TT |
5486 | if (err) |
5487 | lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); | |
7c319d32 AK |
5488 | |
5489 | return err; | |
5490 | } | |
5491 | ||
5492 | /* Enable usage tracking for all quota types. */ | |
5493 | static int ext4_enable_quotas(struct super_block *sb) | |
5494 | { | |
5495 | int type, err = 0; | |
a2d4a646 | 5496 | unsigned long qf_inums[EXT4_MAXQUOTAS] = { |
7c319d32 | 5497 | le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), |
689c958c LX |
5498 | le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), |
5499 | le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) | |
7c319d32 | 5500 | }; |
49da9392 JK |
5501 | bool quota_mopt[EXT4_MAXQUOTAS] = { |
5502 | test_opt(sb, USRQUOTA), | |
5503 | test_opt(sb, GRPQUOTA), | |
5504 | test_opt(sb, PRJQUOTA), | |
5505 | }; | |
7c319d32 AK |
5506 | |
5507 | sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; | |
a2d4a646 | 5508 | for (type = 0; type < EXT4_MAXQUOTAS; type++) { |
7c319d32 AK |
5509 | if (qf_inums[type]) { |
5510 | err = ext4_quota_enable(sb, type, QFMT_VFS_V1, | |
49da9392 JK |
5511 | DQUOT_USAGE_ENABLED | |
5512 | (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); | |
7c319d32 AK |
5513 | if (err) { |
5514 | ext4_warning(sb, | |
72ba7450 TT |
5515 | "Failed to enable quota tracking " |
5516 | "(type=%d, err=%d). Please run " | |
5517 | "e2fsck to fix.", type, err); | |
7c319d32 AK |
5518 | return err; |
5519 | } | |
5520 | } | |
5521 | } | |
5522 | return 0; | |
5523 | } | |
5524 | ||
ca0e05e4 DM |
5525 | static int ext4_quota_off(struct super_block *sb, int type) |
5526 | { | |
21f97697 JK |
5527 | struct inode *inode = sb_dqopt(sb)->files[type]; |
5528 | handle_t *handle; | |
957153fc | 5529 | int err; |
21f97697 | 5530 | |
87009d86 DM |
5531 | /* Force all delayed allocation blocks to be allocated. |
5532 | * Caller already holds s_umount sem */ | |
5533 | if (test_opt(sb, DELALLOC)) | |
ca0e05e4 | 5534 | sync_filesystem(sb); |
ca0e05e4 | 5535 | |
957153fc | 5536 | if (!inode || !igrab(inode)) |
0b268590 AG |
5537 | goto out; |
5538 | ||
957153fc | 5539 | err = dquot_quota_off(sb, type); |
964edf66 | 5540 | if (err || ext4_has_feature_quota(sb)) |
957153fc JK |
5541 | goto out_put; |
5542 | ||
5543 | inode_lock(inode); | |
61a92987 JK |
5544 | /* |
5545 | * Update modification times of quota files when userspace can | |
5546 | * start looking at them. If we fail, we return success anyway since | |
5547 | * this is not a hard failure and quotas are already disabled. | |
5548 | */ | |
9924a92a | 5549 | handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); |
21f97697 | 5550 | if (IS_ERR(handle)) |
957153fc JK |
5551 | goto out_unlock; |
5552 | EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL); | |
5553 | inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); | |
eeca7ea1 | 5554 | inode->i_mtime = inode->i_ctime = current_time(inode); |
21f97697 JK |
5555 | ext4_mark_inode_dirty(handle, inode); |
5556 | ext4_journal_stop(handle); | |
957153fc JK |
5557 | out_unlock: |
5558 | inode_unlock(inode); | |
5559 | out_put: | |
964edf66 | 5560 | lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL); |
957153fc JK |
5561 | iput(inode); |
5562 | return err; | |
21f97697 | 5563 | out: |
ca0e05e4 DM |
5564 | return dquot_quota_off(sb, type); |
5565 | } | |
5566 | ||
ac27a0ec DK |
5567 | /* Read data from quotafile - avoid pagecache and such because we cannot afford |
5568 | * acquiring the locks... As quota files are never truncated and quota code | |
25985edc | 5569 | * itself serializes the operations (and no one else should touch the files) |
ac27a0ec | 5570 | * we don't have to be afraid of races */ |
617ba13b | 5571 | static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, |
ac27a0ec DK |
5572 | size_t len, loff_t off) |
5573 | { | |
5574 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
725d26d3 | 5575 | ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); |
ac27a0ec DK |
5576 | int offset = off & (sb->s_blocksize - 1); |
5577 | int tocopy; | |
5578 | size_t toread; | |
5579 | struct buffer_head *bh; | |
5580 | loff_t i_size = i_size_read(inode); | |
5581 | ||
5582 | if (off > i_size) | |
5583 | return 0; | |
5584 | if (off+len > i_size) | |
5585 | len = i_size-off; | |
5586 | toread = len; | |
5587 | while (toread > 0) { | |
5588 | tocopy = sb->s_blocksize - offset < toread ? | |
5589 | sb->s_blocksize - offset : toread; | |
1c215028 TT |
5590 | bh = ext4_bread(NULL, inode, blk, 0); |
5591 | if (IS_ERR(bh)) | |
5592 | return PTR_ERR(bh); | |
ac27a0ec DK |
5593 | if (!bh) /* A hole? */ |
5594 | memset(data, 0, tocopy); | |
5595 | else | |
5596 | memcpy(data, bh->b_data+offset, tocopy); | |
5597 | brelse(bh); | |
5598 | offset = 0; | |
5599 | toread -= tocopy; | |
5600 | data += tocopy; | |
5601 | blk++; | |
5602 | } | |
5603 | return len; | |
5604 | } | |
5605 | ||
5606 | /* Write to quotafile (we know the transaction is already started and has | |
5607 | * enough credits) */ | |
617ba13b | 5608 | static ssize_t ext4_quota_write(struct super_block *sb, int type, |
ac27a0ec DK |
5609 | const char *data, size_t len, loff_t off) |
5610 | { | |
5611 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
725d26d3 | 5612 | ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); |
1c215028 | 5613 | int err, offset = off & (sb->s_blocksize - 1); |
c5e298ae | 5614 | int retries = 0; |
ac27a0ec DK |
5615 | struct buffer_head *bh; |
5616 | handle_t *handle = journal_current_handle(); | |
5617 | ||
0390131b | 5618 | if (EXT4_SB(sb)->s_journal && !handle) { |
b31e1552 ES |
5619 | ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" |
5620 | " cancelled because transaction is not started", | |
9c3013e9 JK |
5621 | (unsigned long long)off, (unsigned long long)len); |
5622 | return -EIO; | |
5623 | } | |
67eeb568 DM |
5624 | /* |
5625 | * Since we account only one data block in transaction credits, | |
5626 | * then it is impossible to cross a block boundary. | |
5627 | */ | |
5628 | if (sb->s_blocksize - offset < len) { | |
5629 | ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" | |
5630 | " cancelled because not block aligned", | |
5631 | (unsigned long long)off, (unsigned long long)len); | |
5632 | return -EIO; | |
5633 | } | |
5634 | ||
c5e298ae TT |
5635 | do { |
5636 | bh = ext4_bread(handle, inode, blk, | |
5637 | EXT4_GET_BLOCKS_CREATE | | |
5638 | EXT4_GET_BLOCKS_METADATA_NOFAIL); | |
5639 | } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) && | |
5640 | ext4_should_retry_alloc(inode->i_sb, &retries)); | |
1c215028 TT |
5641 | if (IS_ERR(bh)) |
5642 | return PTR_ERR(bh); | |
67eeb568 DM |
5643 | if (!bh) |
5644 | goto out; | |
5d601255 | 5645 | BUFFER_TRACE(bh, "get write access"); |
62d2b5f2 JK |
5646 | err = ext4_journal_get_write_access(handle, bh); |
5647 | if (err) { | |
5648 | brelse(bh); | |
1c215028 | 5649 | return err; |
ac27a0ec | 5650 | } |
67eeb568 DM |
5651 | lock_buffer(bh); |
5652 | memcpy(bh->b_data+offset, data, len); | |
5653 | flush_dcache_page(bh->b_page); | |
5654 | unlock_buffer(bh); | |
62d2b5f2 | 5655 | err = ext4_handle_dirty_metadata(handle, NULL, bh); |
67eeb568 | 5656 | brelse(bh); |
ac27a0ec | 5657 | out: |
67eeb568 DM |
5658 | if (inode->i_size < off + len) { |
5659 | i_size_write(inode, off + len); | |
617ba13b | 5660 | EXT4_I(inode)->i_disksize = inode->i_size; |
21f97697 | 5661 | ext4_mark_inode_dirty(handle, inode); |
ac27a0ec | 5662 | } |
67eeb568 | 5663 | return len; |
ac27a0ec DK |
5664 | } |
5665 | ||
8f0e8746 TT |
5666 | static int ext4_get_next_id(struct super_block *sb, struct kqid *qid) |
5667 | { | |
5668 | const struct quota_format_ops *ops; | |
5669 | ||
5670 | if (!sb_has_quota_loaded(sb, qid->type)) | |
5671 | return -ESRCH; | |
5672 | ops = sb_dqopt(sb)->ops[qid->type]; | |
5673 | if (!ops || !ops->get_next_id) | |
5674 | return -ENOSYS; | |
5675 | return dquot_get_next_id(sb, qid); | |
5676 | } | |
ac27a0ec DK |
5677 | #endif |
5678 | ||
152a0836 AV |
5679 | static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, |
5680 | const char *dev_name, void *data) | |
ac27a0ec | 5681 | { |
152a0836 | 5682 | return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); |
ac27a0ec DK |
5683 | } |
5684 | ||
c290ea01 | 5685 | #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) |
24b58424 TT |
5686 | static inline void register_as_ext2(void) |
5687 | { | |
5688 | int err = register_filesystem(&ext2_fs_type); | |
5689 | if (err) | |
5690 | printk(KERN_WARNING | |
5691 | "EXT4-fs: Unable to register as ext2 (%d)\n", err); | |
5692 | } | |
5693 | ||
5694 | static inline void unregister_as_ext2(void) | |
5695 | { | |
5696 | unregister_filesystem(&ext2_fs_type); | |
5697 | } | |
2035e776 TT |
5698 | |
5699 | static inline int ext2_feature_set_ok(struct super_block *sb) | |
5700 | { | |
e2b911c5 | 5701 | if (ext4_has_unknown_ext2_incompat_features(sb)) |
2035e776 | 5702 | return 0; |
bc98a42c | 5703 | if (sb_rdonly(sb)) |
2035e776 | 5704 | return 1; |
e2b911c5 | 5705 | if (ext4_has_unknown_ext2_ro_compat_features(sb)) |
2035e776 TT |
5706 | return 0; |
5707 | return 1; | |
5708 | } | |
24b58424 TT |
5709 | #else |
5710 | static inline void register_as_ext2(void) { } | |
5711 | static inline void unregister_as_ext2(void) { } | |
2035e776 | 5712 | static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } |
24b58424 TT |
5713 | #endif |
5714 | ||
24b58424 TT |
5715 | static inline void register_as_ext3(void) |
5716 | { | |
5717 | int err = register_filesystem(&ext3_fs_type); | |
5718 | if (err) | |
5719 | printk(KERN_WARNING | |
5720 | "EXT4-fs: Unable to register as ext3 (%d)\n", err); | |
5721 | } | |
5722 | ||
5723 | static inline void unregister_as_ext3(void) | |
5724 | { | |
5725 | unregister_filesystem(&ext3_fs_type); | |
5726 | } | |
2035e776 TT |
5727 | |
5728 | static inline int ext3_feature_set_ok(struct super_block *sb) | |
5729 | { | |
e2b911c5 | 5730 | if (ext4_has_unknown_ext3_incompat_features(sb)) |
2035e776 | 5731 | return 0; |
e2b911c5 | 5732 | if (!ext4_has_feature_journal(sb)) |
2035e776 | 5733 | return 0; |
bc98a42c | 5734 | if (sb_rdonly(sb)) |
2035e776 | 5735 | return 1; |
e2b911c5 | 5736 | if (ext4_has_unknown_ext3_ro_compat_features(sb)) |
2035e776 TT |
5737 | return 0; |
5738 | return 1; | |
5739 | } | |
24b58424 | 5740 | |
03010a33 TT |
5741 | static struct file_system_type ext4_fs_type = { |
5742 | .owner = THIS_MODULE, | |
5743 | .name = "ext4", | |
152a0836 | 5744 | .mount = ext4_mount, |
03010a33 TT |
5745 | .kill_sb = kill_block_super, |
5746 | .fs_flags = FS_REQUIRES_DEV, | |
5747 | }; | |
7f78e035 | 5748 | MODULE_ALIAS_FS("ext4"); |
03010a33 | 5749 | |
e9e3bcec ES |
5750 | /* Shared across all ext4 file systems */ |
5751 | wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; | |
e9e3bcec | 5752 | |
5dabfc78 | 5753 | static int __init ext4_init_fs(void) |
ac27a0ec | 5754 | { |
e9e3bcec | 5755 | int i, err; |
c9de560d | 5756 | |
e294a537 | 5757 | ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); |
07c0c5d8 AV |
5758 | ext4_li_info = NULL; |
5759 | mutex_init(&ext4_li_mtx); | |
5760 | ||
9a4c8019 | 5761 | /* Build-time check for flags consistency */ |
12e9b892 | 5762 | ext4_check_flag_values(); |
e9e3bcec | 5763 | |
e142d052 | 5764 | for (i = 0; i < EXT4_WQ_HASH_SZ; i++) |
e9e3bcec | 5765 | init_waitqueue_head(&ext4__ioend_wq[i]); |
e9e3bcec | 5766 | |
51865fda | 5767 | err = ext4_init_es(); |
6fd058f7 TT |
5768 | if (err) |
5769 | return err; | |
51865fda ZL |
5770 | |
5771 | err = ext4_init_pageio(); | |
5772 | if (err) | |
b5799018 | 5773 | goto out5; |
51865fda | 5774 | |
5dabfc78 | 5775 | err = ext4_init_system_zone(); |
bd2d0210 | 5776 | if (err) |
b5799018 | 5777 | goto out4; |
857ac889 | 5778 | |
b5799018 | 5779 | err = ext4_init_sysfs(); |
dd68314c | 5780 | if (err) |
b5799018 | 5781 | goto out3; |
857ac889 | 5782 | |
5dabfc78 | 5783 | err = ext4_init_mballoc(); |
c9de560d AT |
5784 | if (err) |
5785 | goto out2; | |
ac27a0ec DK |
5786 | err = init_inodecache(); |
5787 | if (err) | |
5788 | goto out1; | |
24b58424 | 5789 | register_as_ext3(); |
2035e776 | 5790 | register_as_ext2(); |
03010a33 | 5791 | err = register_filesystem(&ext4_fs_type); |
ac27a0ec DK |
5792 | if (err) |
5793 | goto out; | |
bfff6873 | 5794 | |
ac27a0ec DK |
5795 | return 0; |
5796 | out: | |
24b58424 TT |
5797 | unregister_as_ext2(); |
5798 | unregister_as_ext3(); | |
ac27a0ec DK |
5799 | destroy_inodecache(); |
5800 | out1: | |
5dabfc78 | 5801 | ext4_exit_mballoc(); |
9c191f70 | 5802 | out2: |
b5799018 TT |
5803 | ext4_exit_sysfs(); |
5804 | out3: | |
5dabfc78 | 5805 | ext4_exit_system_zone(); |
b5799018 | 5806 | out4: |
5dabfc78 | 5807 | ext4_exit_pageio(); |
b5799018 | 5808 | out5: |
51865fda ZL |
5809 | ext4_exit_es(); |
5810 | ||
ac27a0ec DK |
5811 | return err; |
5812 | } | |
5813 | ||
5dabfc78 | 5814 | static void __exit ext4_exit_fs(void) |
ac27a0ec | 5815 | { |
bfff6873 | 5816 | ext4_destroy_lazyinit_thread(); |
24b58424 TT |
5817 | unregister_as_ext2(); |
5818 | unregister_as_ext3(); | |
03010a33 | 5819 | unregister_filesystem(&ext4_fs_type); |
ac27a0ec | 5820 | destroy_inodecache(); |
5dabfc78 | 5821 | ext4_exit_mballoc(); |
b5799018 | 5822 | ext4_exit_sysfs(); |
5dabfc78 TT |
5823 | ext4_exit_system_zone(); |
5824 | ext4_exit_pageio(); | |
dd12ed14 | 5825 | ext4_exit_es(); |
ac27a0ec DK |
5826 | } |
5827 | ||
5828 | MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); | |
83982b6f | 5829 | MODULE_DESCRIPTION("Fourth Extended Filesystem"); |
ac27a0ec | 5830 | MODULE_LICENSE("GPL"); |
5dabfc78 TT |
5831 | module_init(ext4_init_fs) |
5832 | module_exit(ext4_exit_fs) |