]>
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> |
dab291af | 24 | #include <linux/jbd2.h> |
ac27a0ec DK |
25 | #include <linux/slab.h> |
26 | #include <linux/init.h> | |
27 | #include <linux/blkdev.h> | |
28 | #include <linux/parser.h> | |
ac27a0ec | 29 | #include <linux/buffer_head.h> |
a5694255 | 30 | #include <linux/exportfs.h> |
ac27a0ec DK |
31 | #include <linux/vfs.h> |
32 | #include <linux/random.h> | |
33 | #include <linux/mount.h> | |
34 | #include <linux/namei.h> | |
35 | #include <linux/quotaops.h> | |
36 | #include <linux/seq_file.h> | |
9f6200bb | 37 | #include <linux/proc_fs.h> |
3197ebdb | 38 | #include <linux/ctype.h> |
1330593e | 39 | #include <linux/log2.h> |
717d50e4 | 40 | #include <linux/crc16.h> |
ac27a0ec DK |
41 | #include <asm/uaccess.h> |
42 | ||
bfff6873 LC |
43 | #include <linux/kthread.h> |
44 | #include <linux/freezer.h> | |
45 | ||
3dcf5451 CH |
46 | #include "ext4.h" |
47 | #include "ext4_jbd2.h" | |
ac27a0ec DK |
48 | #include "xattr.h" |
49 | #include "acl.h" | |
3661d286 | 50 | #include "mballoc.h" |
ac27a0ec | 51 | |
9bffad1e TT |
52 | #define CREATE_TRACE_POINTS |
53 | #include <trace/events/ext4.h> | |
54 | ||
1f109d5a | 55 | static struct proc_dir_entry *ext4_proc_root; |
3197ebdb | 56 | static struct kset *ext4_kset; |
bfff6873 LC |
57 | struct ext4_lazy_init *ext4_li_info; |
58 | struct mutex ext4_li_mtx; | |
857ac889 | 59 | struct ext4_features *ext4_feat; |
9f6200bb | 60 | |
617ba13b | 61 | static int ext4_load_journal(struct super_block *, struct ext4_super_block *, |
ac27a0ec | 62 | unsigned long journal_devnum); |
e2d67052 | 63 | static int ext4_commit_super(struct super_block *sb, int sync); |
2b2d6d01 TT |
64 | static void ext4_mark_recovery_complete(struct super_block *sb, |
65 | struct ext4_super_block *es); | |
66 | static void ext4_clear_journal_err(struct super_block *sb, | |
67 | struct ext4_super_block *es); | |
617ba13b | 68 | static int ext4_sync_fs(struct super_block *sb, int wait); |
2b2d6d01 | 69 | static const char *ext4_decode_error(struct super_block *sb, int errno, |
ac27a0ec | 70 | char nbuf[16]); |
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); |
2b2d6d01 | 74 | static void ext4_write_super(struct super_block *sb); |
c4be0c1d | 75 | static int ext4_freeze(struct super_block *sb); |
152a0836 AV |
76 | static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, |
77 | const char *dev_name, void *data); | |
bfff6873 LC |
78 | static void ext4_destroy_lazyinit_thread(void); |
79 | static void ext4_unregister_li_request(struct super_block *sb); | |
ac27a0ec | 80 | |
ba69f9ab JK |
81 | #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) |
82 | static struct file_system_type ext3_fs_type = { | |
83 | .owner = THIS_MODULE, | |
84 | .name = "ext3", | |
152a0836 | 85 | .mount = ext4_mount, |
ba69f9ab JK |
86 | .kill_sb = kill_block_super, |
87 | .fs_flags = FS_REQUIRES_DEV, | |
88 | }; | |
89 | #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) | |
90 | #else | |
91 | #define IS_EXT3_SB(sb) (0) | |
92 | #endif | |
bd81d8ee | 93 | |
8fadc143 AR |
94 | ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, |
95 | struct ext4_group_desc *bg) | |
bd81d8ee | 96 | { |
3a14589c | 97 | return le32_to_cpu(bg->bg_block_bitmap_lo) | |
8fadc143 | 98 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 99 | (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); |
bd81d8ee LV |
100 | } |
101 | ||
8fadc143 AR |
102 | ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, |
103 | struct ext4_group_desc *bg) | |
bd81d8ee | 104 | { |
5272f837 | 105 | return le32_to_cpu(bg->bg_inode_bitmap_lo) | |
8fadc143 | 106 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 107 | (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); |
bd81d8ee LV |
108 | } |
109 | ||
8fadc143 AR |
110 | ext4_fsblk_t ext4_inode_table(struct super_block *sb, |
111 | struct ext4_group_desc *bg) | |
bd81d8ee | 112 | { |
5272f837 | 113 | return le32_to_cpu(bg->bg_inode_table_lo) | |
8fadc143 | 114 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 115 | (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); |
bd81d8ee LV |
116 | } |
117 | ||
560671a0 AK |
118 | __u32 ext4_free_blks_count(struct super_block *sb, |
119 | struct ext4_group_desc *bg) | |
120 | { | |
121 | return le16_to_cpu(bg->bg_free_blocks_count_lo) | | |
122 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 123 | (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); |
560671a0 AK |
124 | } |
125 | ||
126 | __u32 ext4_free_inodes_count(struct super_block *sb, | |
127 | struct ext4_group_desc *bg) | |
128 | { | |
129 | return le16_to_cpu(bg->bg_free_inodes_count_lo) | | |
130 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 131 | (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); |
560671a0 AK |
132 | } |
133 | ||
134 | __u32 ext4_used_dirs_count(struct super_block *sb, | |
135 | struct ext4_group_desc *bg) | |
136 | { | |
137 | return le16_to_cpu(bg->bg_used_dirs_count_lo) | | |
138 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 139 | (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); |
560671a0 AK |
140 | } |
141 | ||
142 | __u32 ext4_itable_unused_count(struct super_block *sb, | |
143 | struct ext4_group_desc *bg) | |
144 | { | |
145 | return le16_to_cpu(bg->bg_itable_unused_lo) | | |
146 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 147 | (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); |
560671a0 AK |
148 | } |
149 | ||
8fadc143 AR |
150 | void ext4_block_bitmap_set(struct super_block *sb, |
151 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 152 | { |
3a14589c | 153 | bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
154 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
155 | bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
156 | } |
157 | ||
8fadc143 AR |
158 | void ext4_inode_bitmap_set(struct super_block *sb, |
159 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 160 | { |
5272f837 | 161 | bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
162 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
163 | bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
164 | } |
165 | ||
8fadc143 AR |
166 | void ext4_inode_table_set(struct super_block *sb, |
167 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 168 | { |
5272f837 | 169 | bg->bg_inode_table_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
170 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
171 | bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
172 | } |
173 | ||
560671a0 AK |
174 | void ext4_free_blks_set(struct super_block *sb, |
175 | struct ext4_group_desc *bg, __u32 count) | |
176 | { | |
177 | bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); | |
178 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
179 | bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); | |
180 | } | |
181 | ||
182 | void ext4_free_inodes_set(struct super_block *sb, | |
183 | struct ext4_group_desc *bg, __u32 count) | |
184 | { | |
185 | bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); | |
186 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
187 | bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); | |
188 | } | |
189 | ||
190 | void ext4_used_dirs_set(struct super_block *sb, | |
191 | struct ext4_group_desc *bg, __u32 count) | |
192 | { | |
193 | bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); | |
194 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
195 | bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); | |
196 | } | |
197 | ||
198 | void ext4_itable_unused_set(struct super_block *sb, | |
199 | struct ext4_group_desc *bg, __u32 count) | |
200 | { | |
201 | bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); | |
202 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
203 | bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); | |
204 | } | |
205 | ||
d3d1faf6 CW |
206 | |
207 | /* Just increment the non-pointer handle value */ | |
208 | static handle_t *ext4_get_nojournal(void) | |
209 | { | |
210 | handle_t *handle = current->journal_info; | |
211 | unsigned long ref_cnt = (unsigned long)handle; | |
212 | ||
213 | BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT); | |
214 | ||
215 | ref_cnt++; | |
216 | handle = (handle_t *)ref_cnt; | |
217 | ||
218 | current->journal_info = handle; | |
219 | return handle; | |
220 | } | |
221 | ||
222 | ||
223 | /* Decrement the non-pointer handle value */ | |
224 | static void ext4_put_nojournal(handle_t *handle) | |
225 | { | |
226 | unsigned long ref_cnt = (unsigned long)handle; | |
227 | ||
228 | BUG_ON(ref_cnt == 0); | |
229 | ||
230 | ref_cnt--; | |
231 | handle = (handle_t *)ref_cnt; | |
232 | ||
233 | current->journal_info = handle; | |
234 | } | |
235 | ||
ac27a0ec | 236 | /* |
dab291af | 237 | * Wrappers for jbd2_journal_start/end. |
ac27a0ec DK |
238 | * |
239 | * The only special thing we need to do here is to make sure that all | |
240 | * journal_end calls result in the superblock being marked dirty, so | |
241 | * that sync() will call the filesystem's write_super callback if | |
242 | * appropriate. | |
243 | */ | |
617ba13b | 244 | handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks) |
ac27a0ec DK |
245 | { |
246 | journal_t *journal; | |
247 | ||
248 | if (sb->s_flags & MS_RDONLY) | |
249 | return ERR_PTR(-EROFS); | |
250 | ||
437f88cc | 251 | vfs_check_frozen(sb, SB_FREEZE_TRANS); |
ac27a0ec DK |
252 | /* Special case here: if the journal has aborted behind our |
253 | * backs (eg. EIO in the commit thread), then we still need to | |
254 | * take the FS itself readonly cleanly. */ | |
617ba13b | 255 | journal = EXT4_SB(sb)->s_journal; |
0390131b FM |
256 | if (journal) { |
257 | if (is_journal_aborted(journal)) { | |
c67d859e | 258 | ext4_abort(sb, "Detected aborted journal"); |
0390131b FM |
259 | return ERR_PTR(-EROFS); |
260 | } | |
261 | return jbd2_journal_start(journal, nblocks); | |
ac27a0ec | 262 | } |
d3d1faf6 | 263 | return ext4_get_nojournal(); |
ac27a0ec DK |
264 | } |
265 | ||
266 | /* | |
267 | * The only special thing we need to do here is to make sure that all | |
dab291af | 268 | * jbd2_journal_stop calls result in the superblock being marked dirty, so |
ac27a0ec DK |
269 | * that sync() will call the filesystem's write_super callback if |
270 | * appropriate. | |
271 | */ | |
c398eda0 | 272 | int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle) |
ac27a0ec DK |
273 | { |
274 | struct super_block *sb; | |
275 | int err; | |
276 | int rc; | |
277 | ||
0390131b | 278 | if (!ext4_handle_valid(handle)) { |
d3d1faf6 | 279 | ext4_put_nojournal(handle); |
0390131b FM |
280 | return 0; |
281 | } | |
ac27a0ec DK |
282 | sb = handle->h_transaction->t_journal->j_private; |
283 | err = handle->h_err; | |
dab291af | 284 | rc = jbd2_journal_stop(handle); |
ac27a0ec DK |
285 | |
286 | if (!err) | |
287 | err = rc; | |
288 | if (err) | |
c398eda0 | 289 | __ext4_std_error(sb, where, line, err); |
ac27a0ec DK |
290 | return err; |
291 | } | |
292 | ||
90c7201b TT |
293 | void ext4_journal_abort_handle(const char *caller, unsigned int line, |
294 | const char *err_fn, struct buffer_head *bh, | |
295 | handle_t *handle, int err) | |
ac27a0ec DK |
296 | { |
297 | char nbuf[16]; | |
617ba13b | 298 | const char *errstr = ext4_decode_error(NULL, err, nbuf); |
ac27a0ec | 299 | |
0390131b FM |
300 | BUG_ON(!ext4_handle_valid(handle)); |
301 | ||
ac27a0ec DK |
302 | if (bh) |
303 | BUFFER_TRACE(bh, "abort"); | |
304 | ||
305 | if (!handle->h_err) | |
306 | handle->h_err = err; | |
307 | ||
308 | if (is_handle_aborted(handle)) | |
309 | return; | |
310 | ||
90c7201b TT |
311 | printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n", |
312 | caller, line, errstr, err_fn); | |
ac27a0ec | 313 | |
dab291af | 314 | jbd2_journal_abort_handle(handle); |
ac27a0ec DK |
315 | } |
316 | ||
1c13d5c0 TT |
317 | static void __save_error_info(struct super_block *sb, const char *func, |
318 | unsigned int line) | |
319 | { | |
320 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; | |
321 | ||
322 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; | |
323 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
324 | es->s_last_error_time = cpu_to_le32(get_seconds()); | |
325 | strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func)); | |
326 | es->s_last_error_line = cpu_to_le32(line); | |
327 | if (!es->s_first_error_time) { | |
328 | es->s_first_error_time = es->s_last_error_time; | |
329 | strncpy(es->s_first_error_func, func, | |
330 | sizeof(es->s_first_error_func)); | |
331 | es->s_first_error_line = cpu_to_le32(line); | |
332 | es->s_first_error_ino = es->s_last_error_ino; | |
333 | es->s_first_error_block = es->s_last_error_block; | |
334 | } | |
66e61a9e TT |
335 | /* |
336 | * Start the daily error reporting function if it hasn't been | |
337 | * started already | |
338 | */ | |
339 | if (!es->s_error_count) | |
340 | mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ); | |
1c13d5c0 TT |
341 | es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1); |
342 | } | |
343 | ||
344 | static void save_error_info(struct super_block *sb, const char *func, | |
345 | unsigned int line) | |
346 | { | |
347 | __save_error_info(sb, func, line); | |
348 | ext4_commit_super(sb, 1); | |
349 | } | |
350 | ||
351 | ||
ac27a0ec DK |
352 | /* Deal with the reporting of failure conditions on a filesystem such as |
353 | * inconsistencies detected or read IO failures. | |
354 | * | |
355 | * On ext2, we can store the error state of the filesystem in the | |
617ba13b | 356 | * superblock. That is not possible on ext4, because we may have other |
ac27a0ec DK |
357 | * write ordering constraints on the superblock which prevent us from |
358 | * writing it out straight away; and given that the journal is about to | |
359 | * be aborted, we can't rely on the current, or future, transactions to | |
360 | * write out the superblock safely. | |
361 | * | |
dab291af | 362 | * We'll just use the jbd2_journal_abort() error code to record an error in |
d6b198bc | 363 | * the journal instead. On recovery, the journal will complain about |
ac27a0ec DK |
364 | * that error until we've noted it down and cleared it. |
365 | */ | |
366 | ||
617ba13b | 367 | static void ext4_handle_error(struct super_block *sb) |
ac27a0ec | 368 | { |
ac27a0ec DK |
369 | if (sb->s_flags & MS_RDONLY) |
370 | return; | |
371 | ||
2b2d6d01 | 372 | if (!test_opt(sb, ERRORS_CONT)) { |
617ba13b | 373 | journal_t *journal = EXT4_SB(sb)->s_journal; |
ac27a0ec | 374 | |
4ab2f15b | 375 | EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; |
ac27a0ec | 376 | if (journal) |
dab291af | 377 | jbd2_journal_abort(journal, -EIO); |
ac27a0ec | 378 | } |
2b2d6d01 | 379 | if (test_opt(sb, ERRORS_RO)) { |
b31e1552 | 380 | ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); |
ac27a0ec DK |
381 | sb->s_flags |= MS_RDONLY; |
382 | } | |
ac27a0ec | 383 | if (test_opt(sb, ERRORS_PANIC)) |
617ba13b | 384 | panic("EXT4-fs (device %s): panic forced after error\n", |
ac27a0ec DK |
385 | sb->s_id); |
386 | } | |
387 | ||
12062ddd | 388 | void __ext4_error(struct super_block *sb, const char *function, |
c398eda0 | 389 | unsigned int line, const char *fmt, ...) |
ac27a0ec DK |
390 | { |
391 | va_list args; | |
392 | ||
393 | va_start(args, fmt); | |
c398eda0 TT |
394 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: ", |
395 | sb->s_id, function, line, current->comm); | |
ac27a0ec DK |
396 | vprintk(fmt, args); |
397 | printk("\n"); | |
398 | va_end(args); | |
399 | ||
617ba13b | 400 | ext4_handle_error(sb); |
ac27a0ec DK |
401 | } |
402 | ||
c398eda0 TT |
403 | void ext4_error_inode(struct inode *inode, const char *function, |
404 | unsigned int line, ext4_fsblk_t block, | |
273df556 FM |
405 | const char *fmt, ...) |
406 | { | |
407 | va_list args; | |
1c13d5c0 | 408 | struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es; |
273df556 | 409 | |
1c13d5c0 TT |
410 | es->s_last_error_ino = cpu_to_le32(inode->i_ino); |
411 | es->s_last_error_block = cpu_to_le64(block); | |
412 | save_error_info(inode->i_sb, function, line); | |
273df556 | 413 | va_start(args, fmt); |
c398eda0 TT |
414 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ", |
415 | inode->i_sb->s_id, function, line, inode->i_ino); | |
416 | if (block) | |
417 | printk("block %llu: ", block); | |
418 | printk("comm %s: ", current->comm); | |
273df556 FM |
419 | vprintk(fmt, args); |
420 | printk("\n"); | |
421 | va_end(args); | |
422 | ||
423 | ext4_handle_error(inode->i_sb); | |
424 | } | |
425 | ||
c398eda0 TT |
426 | void ext4_error_file(struct file *file, const char *function, |
427 | unsigned int line, const char *fmt, ...) | |
273df556 FM |
428 | { |
429 | va_list args; | |
1c13d5c0 | 430 | struct ext4_super_block *es; |
273df556 FM |
431 | struct inode *inode = file->f_dentry->d_inode; |
432 | char pathname[80], *path; | |
433 | ||
1c13d5c0 TT |
434 | es = EXT4_SB(inode->i_sb)->s_es; |
435 | es->s_last_error_ino = cpu_to_le32(inode->i_ino); | |
436 | save_error_info(inode->i_sb, function, line); | |
273df556 FM |
437 | va_start(args, fmt); |
438 | path = d_path(&(file->f_path), pathname, sizeof(pathname)); | |
439 | if (!path) | |
440 | path = "(unknown)"; | |
441 | printk(KERN_CRIT | |
c398eda0 TT |
442 | "EXT4-fs error (device %s): %s:%d: inode #%lu " |
443 | "(comm %s path %s): ", | |
444 | inode->i_sb->s_id, function, line, inode->i_ino, | |
445 | current->comm, path); | |
273df556 FM |
446 | vprintk(fmt, args); |
447 | printk("\n"); | |
448 | va_end(args); | |
449 | ||
450 | ext4_handle_error(inode->i_sb); | |
451 | } | |
452 | ||
2b2d6d01 | 453 | static const char *ext4_decode_error(struct super_block *sb, int errno, |
ac27a0ec DK |
454 | char nbuf[16]) |
455 | { | |
456 | char *errstr = NULL; | |
457 | ||
458 | switch (errno) { | |
459 | case -EIO: | |
460 | errstr = "IO failure"; | |
461 | break; | |
462 | case -ENOMEM: | |
463 | errstr = "Out of memory"; | |
464 | break; | |
465 | case -EROFS: | |
78f1ddbb TT |
466 | if (!sb || (EXT4_SB(sb)->s_journal && |
467 | EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) | |
ac27a0ec DK |
468 | errstr = "Journal has aborted"; |
469 | else | |
470 | errstr = "Readonly filesystem"; | |
471 | break; | |
472 | default: | |
473 | /* If the caller passed in an extra buffer for unknown | |
474 | * errors, textualise them now. Else we just return | |
475 | * NULL. */ | |
476 | if (nbuf) { | |
477 | /* Check for truncated error codes... */ | |
478 | if (snprintf(nbuf, 16, "error %d", -errno) >= 0) | |
479 | errstr = nbuf; | |
480 | } | |
481 | break; | |
482 | } | |
483 | ||
484 | return errstr; | |
485 | } | |
486 | ||
617ba13b | 487 | /* __ext4_std_error decodes expected errors from journaling functions |
ac27a0ec DK |
488 | * automatically and invokes the appropriate error response. */ |
489 | ||
c398eda0 TT |
490 | void __ext4_std_error(struct super_block *sb, const char *function, |
491 | unsigned int line, int errno) | |
ac27a0ec DK |
492 | { |
493 | char nbuf[16]; | |
494 | const char *errstr; | |
495 | ||
496 | /* Special case: if the error is EROFS, and we're not already | |
497 | * inside a transaction, then there's really no point in logging | |
498 | * an error. */ | |
499 | if (errno == -EROFS && journal_current_handle() == NULL && | |
500 | (sb->s_flags & MS_RDONLY)) | |
501 | return; | |
502 | ||
617ba13b | 503 | errstr = ext4_decode_error(sb, errno, nbuf); |
c398eda0 TT |
504 | printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", |
505 | sb->s_id, function, line, errstr); | |
1c13d5c0 | 506 | save_error_info(sb, function, line); |
ac27a0ec | 507 | |
617ba13b | 508 | ext4_handle_error(sb); |
ac27a0ec DK |
509 | } |
510 | ||
511 | /* | |
617ba13b | 512 | * ext4_abort is a much stronger failure handler than ext4_error. The |
ac27a0ec DK |
513 | * abort function may be used to deal with unrecoverable failures such |
514 | * as journal IO errors or ENOMEM at a critical moment in log management. | |
515 | * | |
516 | * We unconditionally force the filesystem into an ABORT|READONLY state, | |
517 | * unless the error response on the fs has been set to panic in which | |
518 | * case we take the easy way out and panic immediately. | |
519 | */ | |
520 | ||
c67d859e | 521 | void __ext4_abort(struct super_block *sb, const char *function, |
c398eda0 | 522 | unsigned int line, const char *fmt, ...) |
ac27a0ec DK |
523 | { |
524 | va_list args; | |
525 | ||
1c13d5c0 | 526 | save_error_info(sb, function, line); |
ac27a0ec | 527 | va_start(args, fmt); |
c398eda0 TT |
528 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id, |
529 | function, line); | |
ac27a0ec DK |
530 | vprintk(fmt, args); |
531 | printk("\n"); | |
532 | va_end(args); | |
533 | ||
1c13d5c0 TT |
534 | if ((sb->s_flags & MS_RDONLY) == 0) { |
535 | ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); | |
536 | sb->s_flags |= MS_RDONLY; | |
537 | EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; | |
538 | if (EXT4_SB(sb)->s_journal) | |
539 | jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); | |
540 | save_error_info(sb, function, line); | |
541 | } | |
ac27a0ec | 542 | if (test_opt(sb, ERRORS_PANIC)) |
617ba13b | 543 | panic("EXT4-fs panic from previous error\n"); |
ac27a0ec DK |
544 | } |
545 | ||
b31e1552 ES |
546 | void ext4_msg (struct super_block * sb, const char *prefix, |
547 | const char *fmt, ...) | |
548 | { | |
549 | va_list args; | |
550 | ||
551 | va_start(args, fmt); | |
552 | printk("%sEXT4-fs (%s): ", prefix, sb->s_id); | |
553 | vprintk(fmt, args); | |
554 | printk("\n"); | |
555 | va_end(args); | |
556 | } | |
557 | ||
12062ddd | 558 | void __ext4_warning(struct super_block *sb, const char *function, |
c398eda0 | 559 | unsigned int line, const char *fmt, ...) |
ac27a0ec DK |
560 | { |
561 | va_list args; | |
562 | ||
563 | va_start(args, fmt); | |
c398eda0 TT |
564 | printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: ", |
565 | sb->s_id, function, line); | |
ac27a0ec DK |
566 | vprintk(fmt, args); |
567 | printk("\n"); | |
568 | va_end(args); | |
569 | } | |
570 | ||
e29136f8 TT |
571 | void __ext4_grp_locked_error(const char *function, unsigned int line, |
572 | struct super_block *sb, ext4_group_t grp, | |
573 | unsigned long ino, ext4_fsblk_t block, | |
574 | const char *fmt, ...) | |
5d1b1b3f AK |
575 | __releases(bitlock) |
576 | __acquires(bitlock) | |
577 | { | |
578 | va_list args; | |
579 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; | |
580 | ||
1c13d5c0 TT |
581 | es->s_last_error_ino = cpu_to_le32(ino); |
582 | es->s_last_error_block = cpu_to_le64(block); | |
583 | __save_error_info(sb, function, line); | |
5d1b1b3f | 584 | va_start(args, fmt); |
e29136f8 TT |
585 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u", |
586 | sb->s_id, function, line, grp); | |
587 | if (ino) | |
588 | printk("inode %lu: ", ino); | |
589 | if (block) | |
590 | printk("block %llu:", (unsigned long long) block); | |
5d1b1b3f AK |
591 | vprintk(fmt, args); |
592 | printk("\n"); | |
593 | va_end(args); | |
594 | ||
595 | if (test_opt(sb, ERRORS_CONT)) { | |
e2d67052 | 596 | ext4_commit_super(sb, 0); |
5d1b1b3f AK |
597 | return; |
598 | } | |
1c13d5c0 | 599 | |
5d1b1b3f AK |
600 | ext4_unlock_group(sb, grp); |
601 | ext4_handle_error(sb); | |
602 | /* | |
603 | * We only get here in the ERRORS_RO case; relocking the group | |
604 | * may be dangerous, but nothing bad will happen since the | |
605 | * filesystem will have already been marked read/only and the | |
606 | * journal has been aborted. We return 1 as a hint to callers | |
607 | * who might what to use the return value from | |
608 | * ext4_grp_locked_error() to distinguish beween the | |
609 | * ERRORS_CONT and ERRORS_RO case, and perhaps return more | |
610 | * aggressively from the ext4 function in question, with a | |
611 | * more appropriate error code. | |
612 | */ | |
613 | ext4_lock_group(sb, grp); | |
614 | return; | |
615 | } | |
616 | ||
617ba13b | 617 | void ext4_update_dynamic_rev(struct super_block *sb) |
ac27a0ec | 618 | { |
617ba13b | 619 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
ac27a0ec | 620 | |
617ba13b | 621 | if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) |
ac27a0ec DK |
622 | return; |
623 | ||
12062ddd | 624 | ext4_warning(sb, |
ac27a0ec DK |
625 | "updating to rev %d because of new feature flag, " |
626 | "running e2fsck is recommended", | |
617ba13b | 627 | EXT4_DYNAMIC_REV); |
ac27a0ec | 628 | |
617ba13b MC |
629 | es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); |
630 | es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); | |
631 | es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); | |
ac27a0ec DK |
632 | /* leave es->s_feature_*compat flags alone */ |
633 | /* es->s_uuid will be set by e2fsck if empty */ | |
634 | ||
635 | /* | |
636 | * The rest of the superblock fields should be zero, and if not it | |
637 | * means they are likely already in use, so leave them alone. We | |
638 | * can leave it up to e2fsck to clean up any inconsistencies there. | |
639 | */ | |
640 | } | |
641 | ||
642 | /* | |
643 | * Open the external journal device | |
644 | */ | |
b31e1552 | 645 | static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) |
ac27a0ec DK |
646 | { |
647 | struct block_device *bdev; | |
648 | char b[BDEVNAME_SIZE]; | |
649 | ||
650 | bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE); | |
651 | if (IS_ERR(bdev)) | |
652 | goto fail; | |
653 | return bdev; | |
654 | ||
655 | fail: | |
b31e1552 | 656 | ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld", |
ac27a0ec DK |
657 | __bdevname(dev, b), PTR_ERR(bdev)); |
658 | return NULL; | |
659 | } | |
660 | ||
661 | /* | |
662 | * Release the journal device | |
663 | */ | |
617ba13b | 664 | static int ext4_blkdev_put(struct block_device *bdev) |
ac27a0ec DK |
665 | { |
666 | bd_release(bdev); | |
9a1c3542 | 667 | return blkdev_put(bdev, FMODE_READ|FMODE_WRITE); |
ac27a0ec DK |
668 | } |
669 | ||
617ba13b | 670 | static int ext4_blkdev_remove(struct ext4_sb_info *sbi) |
ac27a0ec DK |
671 | { |
672 | struct block_device *bdev; | |
673 | int ret = -ENODEV; | |
674 | ||
675 | bdev = sbi->journal_bdev; | |
676 | if (bdev) { | |
617ba13b | 677 | ret = ext4_blkdev_put(bdev); |
ac27a0ec DK |
678 | sbi->journal_bdev = NULL; |
679 | } | |
680 | return ret; | |
681 | } | |
682 | ||
683 | static inline struct inode *orphan_list_entry(struct list_head *l) | |
684 | { | |
617ba13b | 685 | return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; |
ac27a0ec DK |
686 | } |
687 | ||
617ba13b | 688 | static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) |
ac27a0ec DK |
689 | { |
690 | struct list_head *l; | |
691 | ||
b31e1552 ES |
692 | ext4_msg(sb, KERN_ERR, "sb orphan head is %d", |
693 | le32_to_cpu(sbi->s_es->s_last_orphan)); | |
ac27a0ec DK |
694 | |
695 | printk(KERN_ERR "sb_info orphan list:\n"); | |
696 | list_for_each(l, &sbi->s_orphan) { | |
697 | struct inode *inode = orphan_list_entry(l); | |
698 | printk(KERN_ERR " " | |
699 | "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", | |
700 | inode->i_sb->s_id, inode->i_ino, inode, | |
701 | inode->i_mode, inode->i_nlink, | |
702 | NEXT_ORPHAN(inode)); | |
703 | } | |
704 | } | |
705 | ||
2b2d6d01 | 706 | static void ext4_put_super(struct super_block *sb) |
ac27a0ec | 707 | { |
617ba13b MC |
708 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
709 | struct ext4_super_block *es = sbi->s_es; | |
ef2cabf7 | 710 | int i, err; |
ac27a0ec | 711 | |
857ac889 | 712 | ext4_unregister_li_request(sb); |
e0ccfd95 CH |
713 | dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED); |
714 | ||
4c0425ff MC |
715 | flush_workqueue(sbi->dio_unwritten_wq); |
716 | destroy_workqueue(sbi->dio_unwritten_wq); | |
717 | ||
a9e220f8 | 718 | lock_super(sb); |
8c85e125 | 719 | if (sb->s_dirt) |
ebc1ac16 | 720 | ext4_commit_super(sb, 1); |
8c85e125 | 721 | |
0390131b FM |
722 | if (sbi->s_journal) { |
723 | err = jbd2_journal_destroy(sbi->s_journal); | |
724 | sbi->s_journal = NULL; | |
725 | if (err < 0) | |
c67d859e | 726 | ext4_abort(sb, "Couldn't clean up the journal"); |
0390131b | 727 | } |
d4edac31 | 728 | |
a1c6c569 | 729 | del_timer(&sbi->s_err_report); |
d4edac31 JB |
730 | ext4_release_system_zone(sb); |
731 | ext4_mb_release(sb); | |
732 | ext4_ext_release(sb); | |
733 | ext4_xattr_put_super(sb); | |
734 | ||
ac27a0ec | 735 | if (!(sb->s_flags & MS_RDONLY)) { |
617ba13b | 736 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); |
ac27a0ec | 737 | es->s_state = cpu_to_le16(sbi->s_mount_state); |
e2d67052 | 738 | ext4_commit_super(sb, 1); |
ac27a0ec | 739 | } |
240799cd | 740 | if (sbi->s_proc) { |
9f6200bb | 741 | remove_proc_entry(sb->s_id, ext4_proc_root); |
240799cd | 742 | } |
3197ebdb | 743 | kobject_del(&sbi->s_kobj); |
ac27a0ec DK |
744 | |
745 | for (i = 0; i < sbi->s_gdb_count; i++) | |
746 | brelse(sbi->s_group_desc[i]); | |
747 | kfree(sbi->s_group_desc); | |
c5ca7c76 TT |
748 | if (is_vmalloc_addr(sbi->s_flex_groups)) |
749 | vfree(sbi->s_flex_groups); | |
750 | else | |
751 | kfree(sbi->s_flex_groups); | |
ac27a0ec DK |
752 | percpu_counter_destroy(&sbi->s_freeblocks_counter); |
753 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
754 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
6bc6e63f | 755 | percpu_counter_destroy(&sbi->s_dirtyblocks_counter); |
ac27a0ec DK |
756 | brelse(sbi->s_sbh); |
757 | #ifdef CONFIG_QUOTA | |
758 | for (i = 0; i < MAXQUOTAS; i++) | |
759 | kfree(sbi->s_qf_names[i]); | |
760 | #endif | |
761 | ||
762 | /* Debugging code just in case the in-memory inode orphan list | |
763 | * isn't empty. The on-disk one can be non-empty if we've | |
764 | * detected an error and taken the fs readonly, but the | |
765 | * in-memory list had better be clean by this point. */ | |
766 | if (!list_empty(&sbi->s_orphan)) | |
767 | dump_orphan_list(sb, sbi); | |
768 | J_ASSERT(list_empty(&sbi->s_orphan)); | |
769 | ||
f98393a6 | 770 | invalidate_bdev(sb->s_bdev); |
ac27a0ec DK |
771 | if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { |
772 | /* | |
773 | * Invalidate the journal device's buffers. We don't want them | |
774 | * floating about in memory - the physical journal device may | |
775 | * hotswapped, and it breaks the `ro-after' testing code. | |
776 | */ | |
777 | sync_blockdev(sbi->journal_bdev); | |
f98393a6 | 778 | invalidate_bdev(sbi->journal_bdev); |
617ba13b | 779 | ext4_blkdev_remove(sbi); |
ac27a0ec DK |
780 | } |
781 | sb->s_fs_info = NULL; | |
3197ebdb TT |
782 | /* |
783 | * Now that we are completely done shutting down the | |
784 | * superblock, we need to actually destroy the kobject. | |
785 | */ | |
3197ebdb TT |
786 | unlock_super(sb); |
787 | kobject_put(&sbi->s_kobj); | |
788 | wait_for_completion(&sbi->s_kobj_unregister); | |
705895b6 | 789 | kfree(sbi->s_blockgroup_lock); |
ac27a0ec | 790 | kfree(sbi); |
ac27a0ec DK |
791 | } |
792 | ||
e18b890b | 793 | static struct kmem_cache *ext4_inode_cachep; |
ac27a0ec DK |
794 | |
795 | /* | |
796 | * Called inside transaction, so use GFP_NOFS | |
797 | */ | |
617ba13b | 798 | static struct inode *ext4_alloc_inode(struct super_block *sb) |
ac27a0ec | 799 | { |
617ba13b | 800 | struct ext4_inode_info *ei; |
ac27a0ec | 801 | |
e6b4f8da | 802 | ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); |
ac27a0ec DK |
803 | if (!ei) |
804 | return NULL; | |
0b8e58a1 | 805 | |
ac27a0ec | 806 | ei->vfs_inode.i_version = 1; |
91246c00 | 807 | ei->vfs_inode.i_data.writeback_index = 0; |
a86c6181 | 808 | memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache)); |
c9de560d AT |
809 | INIT_LIST_HEAD(&ei->i_prealloc_list); |
810 | spin_lock_init(&ei->i_prealloc_lock); | |
0390131b FM |
811 | /* |
812 | * Note: We can be called before EXT4_SB(sb)->s_journal is set, | |
813 | * therefore it can be null here. Don't check it, just initialize | |
814 | * jinode. | |
815 | */ | |
678aaf48 | 816 | jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode); |
d2a17637 MC |
817 | ei->i_reserved_data_blocks = 0; |
818 | ei->i_reserved_meta_blocks = 0; | |
819 | ei->i_allocated_meta_blocks = 0; | |
9d0be502 | 820 | ei->i_da_metadata_calc_len = 0; |
d2a17637 MC |
821 | ei->i_delalloc_reserved_flag = 0; |
822 | spin_lock_init(&(ei->i_block_reservation_lock)); | |
a9e7f447 DM |
823 | #ifdef CONFIG_QUOTA |
824 | ei->i_reserved_quota = 0; | |
825 | #endif | |
c7064ef1 | 826 | INIT_LIST_HEAD(&ei->i_completed_io_list); |
744692dc | 827 | spin_lock_init(&ei->i_completed_io_lock); |
8d5d02e6 | 828 | ei->cur_aio_dio = NULL; |
b436b9be JK |
829 | ei->i_sync_tid = 0; |
830 | ei->i_datasync_tid = 0; | |
f7ad6d2e | 831 | atomic_set(&ei->i_ioend_count, 0); |
0b8e58a1 | 832 | |
ac27a0ec DK |
833 | return &ei->vfs_inode; |
834 | } | |
835 | ||
7ff9c073 TT |
836 | static int ext4_drop_inode(struct inode *inode) |
837 | { | |
838 | int drop = generic_drop_inode(inode); | |
839 | ||
840 | trace_ext4_drop_inode(inode, drop); | |
841 | return drop; | |
842 | } | |
843 | ||
617ba13b | 844 | static void ext4_destroy_inode(struct inode *inode) |
ac27a0ec | 845 | { |
f7ad6d2e | 846 | ext4_ioend_wait(inode); |
9f7dd93d | 847 | if (!list_empty(&(EXT4_I(inode)->i_orphan))) { |
b31e1552 ES |
848 | ext4_msg(inode->i_sb, KERN_ERR, |
849 | "Inode %lu (%p): orphan list check failed!", | |
850 | inode->i_ino, EXT4_I(inode)); | |
9f7dd93d VA |
851 | print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, |
852 | EXT4_I(inode), sizeof(struct ext4_inode_info), | |
853 | true); | |
854 | dump_stack(); | |
855 | } | |
617ba13b | 856 | kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); |
ac27a0ec DK |
857 | } |
858 | ||
51cc5068 | 859 | static void init_once(void *foo) |
ac27a0ec | 860 | { |
617ba13b | 861 | struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; |
ac27a0ec | 862 | |
a35afb83 | 863 | INIT_LIST_HEAD(&ei->i_orphan); |
03010a33 | 864 | #ifdef CONFIG_EXT4_FS_XATTR |
a35afb83 | 865 | init_rwsem(&ei->xattr_sem); |
ac27a0ec | 866 | #endif |
0e855ac8 | 867 | init_rwsem(&ei->i_data_sem); |
a35afb83 | 868 | inode_init_once(&ei->vfs_inode); |
ac27a0ec DK |
869 | } |
870 | ||
871 | static int init_inodecache(void) | |
872 | { | |
617ba13b MC |
873 | ext4_inode_cachep = kmem_cache_create("ext4_inode_cache", |
874 | sizeof(struct ext4_inode_info), | |
ac27a0ec DK |
875 | 0, (SLAB_RECLAIM_ACCOUNT| |
876 | SLAB_MEM_SPREAD), | |
20c2df83 | 877 | init_once); |
617ba13b | 878 | if (ext4_inode_cachep == NULL) |
ac27a0ec DK |
879 | return -ENOMEM; |
880 | return 0; | |
881 | } | |
882 | ||
883 | static void destroy_inodecache(void) | |
884 | { | |
617ba13b | 885 | kmem_cache_destroy(ext4_inode_cachep); |
ac27a0ec DK |
886 | } |
887 | ||
0930fcc1 | 888 | void ext4_clear_inode(struct inode *inode) |
ac27a0ec | 889 | { |
0930fcc1 AV |
890 | invalidate_inode_buffers(inode); |
891 | end_writeback(inode); | |
9f754758 | 892 | dquot_drop(inode); |
c2ea3fde | 893 | ext4_discard_preallocations(inode); |
0390131b FM |
894 | if (EXT4_JOURNAL(inode)) |
895 | jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal, | |
678aaf48 | 896 | &EXT4_I(inode)->jinode); |
ac27a0ec DK |
897 | } |
898 | ||
2b2d6d01 TT |
899 | static inline void ext4_show_quota_options(struct seq_file *seq, |
900 | struct super_block *sb) | |
ac27a0ec DK |
901 | { |
902 | #if defined(CONFIG_QUOTA) | |
617ba13b | 903 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 904 | |
5a20bdfc JK |
905 | if (sbi->s_jquota_fmt) { |
906 | char *fmtname = ""; | |
907 | ||
908 | switch (sbi->s_jquota_fmt) { | |
909 | case QFMT_VFS_OLD: | |
910 | fmtname = "vfsold"; | |
911 | break; | |
912 | case QFMT_VFS_V0: | |
913 | fmtname = "vfsv0"; | |
914 | break; | |
915 | case QFMT_VFS_V1: | |
916 | fmtname = "vfsv1"; | |
917 | break; | |
918 | } | |
919 | seq_printf(seq, ",jqfmt=%s", fmtname); | |
920 | } | |
ac27a0ec DK |
921 | |
922 | if (sbi->s_qf_names[USRQUOTA]) | |
923 | seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]); | |
924 | ||
925 | if (sbi->s_qf_names[GRPQUOTA]) | |
926 | seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]); | |
927 | ||
482a7425 | 928 | if (test_opt(sb, USRQUOTA)) |
ac27a0ec DK |
929 | seq_puts(seq, ",usrquota"); |
930 | ||
482a7425 | 931 | if (test_opt(sb, GRPQUOTA)) |
ac27a0ec DK |
932 | seq_puts(seq, ",grpquota"); |
933 | #endif | |
934 | } | |
935 | ||
d9c9bef1 MS |
936 | /* |
937 | * Show an option if | |
938 | * - it's set to a non-default value OR | |
939 | * - if the per-sb default is different from the global default | |
940 | */ | |
617ba13b | 941 | static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs) |
ac27a0ec | 942 | { |
aa22df2c AK |
943 | int def_errors; |
944 | unsigned long def_mount_opts; | |
ac27a0ec | 945 | struct super_block *sb = vfs->mnt_sb; |
d9c9bef1 MS |
946 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
947 | struct ext4_super_block *es = sbi->s_es; | |
d9c9bef1 MS |
948 | |
949 | def_mount_opts = le32_to_cpu(es->s_default_mount_opts); | |
aa22df2c | 950 | def_errors = le16_to_cpu(es->s_errors); |
d9c9bef1 MS |
951 | |
952 | if (sbi->s_sb_block != 1) | |
953 | seq_printf(seq, ",sb=%llu", sbi->s_sb_block); | |
954 | if (test_opt(sb, MINIX_DF)) | |
955 | seq_puts(seq, ",minixdf"); | |
aa22df2c | 956 | if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS)) |
d9c9bef1 MS |
957 | seq_puts(seq, ",grpid"); |
958 | if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS)) | |
959 | seq_puts(seq, ",nogrpid"); | |
960 | if (sbi->s_resuid != EXT4_DEF_RESUID || | |
961 | le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) { | |
962 | seq_printf(seq, ",resuid=%u", sbi->s_resuid); | |
963 | } | |
964 | if (sbi->s_resgid != EXT4_DEF_RESGID || | |
965 | le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) { | |
966 | seq_printf(seq, ",resgid=%u", sbi->s_resgid); | |
967 | } | |
bb4f397a | 968 | if (test_opt(sb, ERRORS_RO)) { |
d9c9bef1 | 969 | if (def_errors == EXT4_ERRORS_PANIC || |
bb4f397a AK |
970 | def_errors == EXT4_ERRORS_CONTINUE) { |
971 | seq_puts(seq, ",errors=remount-ro"); | |
d9c9bef1 MS |
972 | } |
973 | } | |
aa22df2c | 974 | if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) |
bb4f397a | 975 | seq_puts(seq, ",errors=continue"); |
aa22df2c | 976 | if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) |
d9c9bef1 | 977 | seq_puts(seq, ",errors=panic"); |
aa22df2c | 978 | if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16)) |
d9c9bef1 | 979 | seq_puts(seq, ",nouid32"); |
aa22df2c | 980 | if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG)) |
d9c9bef1 MS |
981 | seq_puts(seq, ",debug"); |
982 | if (test_opt(sb, OLDALLOC)) | |
983 | seq_puts(seq, ",oldalloc"); | |
03010a33 | 984 | #ifdef CONFIG_EXT4_FS_XATTR |
aa22df2c AK |
985 | if (test_opt(sb, XATTR_USER) && |
986 | !(def_mount_opts & EXT4_DEFM_XATTR_USER)) | |
d9c9bef1 MS |
987 | seq_puts(seq, ",user_xattr"); |
988 | if (!test_opt(sb, XATTR_USER) && | |
989 | (def_mount_opts & EXT4_DEFM_XATTR_USER)) { | |
990 | seq_puts(seq, ",nouser_xattr"); | |
991 | } | |
992 | #endif | |
03010a33 | 993 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
aa22df2c | 994 | if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL)) |
d9c9bef1 MS |
995 | seq_puts(seq, ",acl"); |
996 | if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL)) | |
997 | seq_puts(seq, ",noacl"); | |
998 | #endif | |
30773840 | 999 | if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { |
d9c9bef1 MS |
1000 | seq_printf(seq, ",commit=%u", |
1001 | (unsigned) (sbi->s_commit_interval / HZ)); | |
1002 | } | |
30773840 TT |
1003 | if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) { |
1004 | seq_printf(seq, ",min_batch_time=%u", | |
1005 | (unsigned) sbi->s_min_batch_time); | |
1006 | } | |
1007 | if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) { | |
1008 | seq_printf(seq, ",max_batch_time=%u", | |
1009 | (unsigned) sbi->s_min_batch_time); | |
1010 | } | |
1011 | ||
571640ca ES |
1012 | /* |
1013 | * We're changing the default of barrier mount option, so | |
1014 | * let's always display its mount state so it's clear what its | |
1015 | * status is. | |
1016 | */ | |
1017 | seq_puts(seq, ",barrier="); | |
1018 | seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0"); | |
cd0b6a39 TT |
1019 | if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) |
1020 | seq_puts(seq, ",journal_async_commit"); | |
39a4bade JK |
1021 | else if (test_opt(sb, JOURNAL_CHECKSUM)) |
1022 | seq_puts(seq, ",journal_checksum"); | |
25ec56b5 JNC |
1023 | if (test_opt(sb, I_VERSION)) |
1024 | seq_puts(seq, ",i_version"); | |
8b67f04a TT |
1025 | if (!test_opt(sb, DELALLOC) && |
1026 | !(def_mount_opts & EXT4_DEFM_NODELALLOC)) | |
dd919b98 AK |
1027 | seq_puts(seq, ",nodelalloc"); |
1028 | ||
1449032b TT |
1029 | if (test_opt(sb, MBLK_IO_SUBMIT)) |
1030 | seq_puts(seq, ",mblk_io_submit"); | |
cb45bbe4 MS |
1031 | if (sbi->s_stripe) |
1032 | seq_printf(seq, ",stripe=%lu", sbi->s_stripe); | |
aa22df2c AK |
1033 | /* |
1034 | * journal mode get enabled in different ways | |
1035 | * So just print the value even if we didn't specify it | |
1036 | */ | |
617ba13b | 1037 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) |
ac27a0ec | 1038 | seq_puts(seq, ",data=journal"); |
617ba13b | 1039 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) |
ac27a0ec | 1040 | seq_puts(seq, ",data=ordered"); |
617ba13b | 1041 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) |
ac27a0ec DK |
1042 | seq_puts(seq, ",data=writeback"); |
1043 | ||
240799cd TT |
1044 | if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) |
1045 | seq_printf(seq, ",inode_readahead_blks=%u", | |
1046 | sbi->s_inode_readahead_blks); | |
1047 | ||
5bf5683a HK |
1048 | if (test_opt(sb, DATA_ERR_ABORT)) |
1049 | seq_puts(seq, ",data_err=abort"); | |
1050 | ||
afd4672d | 1051 | if (test_opt(sb, NO_AUTO_DA_ALLOC)) |
06705bff | 1052 | seq_puts(seq, ",noauto_da_alloc"); |
afd4672d | 1053 | |
8b67f04a | 1054 | if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD)) |
5328e635 ES |
1055 | seq_puts(seq, ",discard"); |
1056 | ||
e3bb52ae ES |
1057 | if (test_opt(sb, NOLOAD)) |
1058 | seq_puts(seq, ",norecovery"); | |
1059 | ||
744692dc JZ |
1060 | if (test_opt(sb, DIOREAD_NOLOCK)) |
1061 | seq_puts(seq, ",dioread_nolock"); | |
1062 | ||
8b67f04a TT |
1063 | if (test_opt(sb, BLOCK_VALIDITY) && |
1064 | !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)) | |
1065 | seq_puts(seq, ",block_validity"); | |
1066 | ||
bfff6873 LC |
1067 | if (!test_opt(sb, INIT_INODE_TABLE)) |
1068 | seq_puts(seq, ",noinit_inode_table"); | |
1069 | else if (sbi->s_li_wait_mult) | |
1070 | seq_printf(seq, ",init_inode_table=%u", | |
1071 | (unsigned) sbi->s_li_wait_mult); | |
1072 | ||
617ba13b | 1073 | ext4_show_quota_options(seq, sb); |
0b8e58a1 | 1074 | |
ac27a0ec DK |
1075 | return 0; |
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 | */ | |
1d1fe1ee DH |
1096 | inode = ext4_iget(sb, ino); |
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, | |
1137 | wait & ~__GFP_WAIT); | |
1138 | return try_to_free_buffers(page); | |
1139 | } | |
1140 | ||
ac27a0ec | 1141 | #ifdef CONFIG_QUOTA |
af5bc92d | 1142 | #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group") |
2b2d6d01 | 1143 | #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA)) |
ac27a0ec | 1144 | |
617ba13b MC |
1145 | static int ext4_write_dquot(struct dquot *dquot); |
1146 | static int ext4_acquire_dquot(struct dquot *dquot); | |
1147 | static int ext4_release_dquot(struct dquot *dquot); | |
1148 | static int ext4_mark_dquot_dirty(struct dquot *dquot); | |
1149 | static int ext4_write_info(struct super_block *sb, int type); | |
6f28e087 | 1150 | static int ext4_quota_on(struct super_block *sb, int type, int format_id, |
307ae18a | 1151 | char *path); |
ca0e05e4 | 1152 | static int ext4_quota_off(struct super_block *sb, int type); |
617ba13b MC |
1153 | static int ext4_quota_on_mount(struct super_block *sb, int type); |
1154 | static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, | |
ac27a0ec | 1155 | size_t len, loff_t off); |
617ba13b | 1156 | static ssize_t ext4_quota_write(struct super_block *sb, int type, |
ac27a0ec DK |
1157 | const char *data, size_t len, loff_t off); |
1158 | ||
61e225dc | 1159 | static const struct dquot_operations ext4_quota_operations = { |
a9e7f447 | 1160 | #ifdef CONFIG_QUOTA |
60e58e0f | 1161 | .get_reserved_space = ext4_get_reserved_space, |
a9e7f447 | 1162 | #endif |
617ba13b MC |
1163 | .write_dquot = ext4_write_dquot, |
1164 | .acquire_dquot = ext4_acquire_dquot, | |
1165 | .release_dquot = ext4_release_dquot, | |
1166 | .mark_dirty = ext4_mark_dquot_dirty, | |
a5b5ee32 JK |
1167 | .write_info = ext4_write_info, |
1168 | .alloc_dquot = dquot_alloc, | |
1169 | .destroy_dquot = dquot_destroy, | |
ac27a0ec DK |
1170 | }; |
1171 | ||
0d54b217 | 1172 | static const struct quotactl_ops ext4_qctl_operations = { |
617ba13b | 1173 | .quota_on = ext4_quota_on, |
ca0e05e4 | 1174 | .quota_off = ext4_quota_off, |
287a8095 CH |
1175 | .quota_sync = dquot_quota_sync, |
1176 | .get_info = dquot_get_dqinfo, | |
1177 | .set_info = dquot_set_dqinfo, | |
1178 | .get_dqblk = dquot_get_dqblk, | |
1179 | .set_dqblk = dquot_set_dqblk | |
ac27a0ec DK |
1180 | }; |
1181 | #endif | |
1182 | ||
ee9b6d61 | 1183 | static const struct super_operations ext4_sops = { |
617ba13b MC |
1184 | .alloc_inode = ext4_alloc_inode, |
1185 | .destroy_inode = ext4_destroy_inode, | |
617ba13b MC |
1186 | .write_inode = ext4_write_inode, |
1187 | .dirty_inode = ext4_dirty_inode, | |
7ff9c073 | 1188 | .drop_inode = ext4_drop_inode, |
0930fcc1 | 1189 | .evict_inode = ext4_evict_inode, |
617ba13b | 1190 | .put_super = ext4_put_super, |
617ba13b | 1191 | .sync_fs = ext4_sync_fs, |
c4be0c1d TS |
1192 | .freeze_fs = ext4_freeze, |
1193 | .unfreeze_fs = ext4_unfreeze, | |
617ba13b MC |
1194 | .statfs = ext4_statfs, |
1195 | .remount_fs = ext4_remount, | |
617ba13b | 1196 | .show_options = ext4_show_options, |
ac27a0ec | 1197 | #ifdef CONFIG_QUOTA |
617ba13b MC |
1198 | .quota_read = ext4_quota_read, |
1199 | .quota_write = ext4_quota_write, | |
ac27a0ec | 1200 | #endif |
c39a7f84 | 1201 | .bdev_try_to_free_page = bdev_try_to_free_page, |
ac27a0ec DK |
1202 | }; |
1203 | ||
9ca92389 TT |
1204 | static const struct super_operations ext4_nojournal_sops = { |
1205 | .alloc_inode = ext4_alloc_inode, | |
1206 | .destroy_inode = ext4_destroy_inode, | |
1207 | .write_inode = ext4_write_inode, | |
1208 | .dirty_inode = ext4_dirty_inode, | |
7ff9c073 | 1209 | .drop_inode = ext4_drop_inode, |
0930fcc1 | 1210 | .evict_inode = ext4_evict_inode, |
9ca92389 TT |
1211 | .write_super = ext4_write_super, |
1212 | .put_super = ext4_put_super, | |
1213 | .statfs = ext4_statfs, | |
1214 | .remount_fs = ext4_remount, | |
9ca92389 TT |
1215 | .show_options = ext4_show_options, |
1216 | #ifdef CONFIG_QUOTA | |
1217 | .quota_read = ext4_quota_read, | |
1218 | .quota_write = ext4_quota_write, | |
1219 | #endif | |
1220 | .bdev_try_to_free_page = bdev_try_to_free_page, | |
1221 | }; | |
1222 | ||
39655164 | 1223 | static const struct export_operations ext4_export_ops = { |
1b961ac0 CH |
1224 | .fh_to_dentry = ext4_fh_to_dentry, |
1225 | .fh_to_parent = ext4_fh_to_parent, | |
617ba13b | 1226 | .get_parent = ext4_get_parent, |
ac27a0ec DK |
1227 | }; |
1228 | ||
1229 | enum { | |
1230 | Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, | |
1231 | Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, | |
01436ef2 | 1232 | Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov, |
ac27a0ec | 1233 | Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, |
06705bff | 1234 | Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh, |
30773840 | 1235 | Opt_commit, Opt_min_batch_time, Opt_max_batch_time, |
c3191067 | 1236 | Opt_journal_update, Opt_journal_dev, |
818d276c | 1237 | Opt_journal_checksum, Opt_journal_async_commit, |
ac27a0ec | 1238 | Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, |
296c355c | 1239 | Opt_data_err_abort, Opt_data_err_ignore, |
ac27a0ec | 1240 | Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, |
5a20bdfc JK |
1241 | Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, |
1242 | Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, | |
1243 | Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version, | |
1449032b TT |
1244 | Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit, |
1245 | Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, | |
5328e635 | 1246 | Opt_inode_readahead_blks, Opt_journal_ioprio, |
744692dc | 1247 | Opt_dioread_nolock, Opt_dioread_lock, |
5328e635 | 1248 | Opt_discard, Opt_nodiscard, |
bfff6873 | 1249 | Opt_init_inode_table, Opt_noinit_inode_table, |
ac27a0ec DK |
1250 | }; |
1251 | ||
a447c093 | 1252 | static const match_table_t tokens = { |
ac27a0ec DK |
1253 | {Opt_bsd_df, "bsddf"}, |
1254 | {Opt_minix_df, "minixdf"}, | |
1255 | {Opt_grpid, "grpid"}, | |
1256 | {Opt_grpid, "bsdgroups"}, | |
1257 | {Opt_nogrpid, "nogrpid"}, | |
1258 | {Opt_nogrpid, "sysvgroups"}, | |
1259 | {Opt_resgid, "resgid=%u"}, | |
1260 | {Opt_resuid, "resuid=%u"}, | |
1261 | {Opt_sb, "sb=%u"}, | |
1262 | {Opt_err_cont, "errors=continue"}, | |
1263 | {Opt_err_panic, "errors=panic"}, | |
1264 | {Opt_err_ro, "errors=remount-ro"}, | |
1265 | {Opt_nouid32, "nouid32"}, | |
ac27a0ec DK |
1266 | {Opt_debug, "debug"}, |
1267 | {Opt_oldalloc, "oldalloc"}, | |
1268 | {Opt_orlov, "orlov"}, | |
1269 | {Opt_user_xattr, "user_xattr"}, | |
1270 | {Opt_nouser_xattr, "nouser_xattr"}, | |
1271 | {Opt_acl, "acl"}, | |
1272 | {Opt_noacl, "noacl"}, | |
ac27a0ec | 1273 | {Opt_noload, "noload"}, |
e3bb52ae | 1274 | {Opt_noload, "norecovery"}, |
ac27a0ec DK |
1275 | {Opt_nobh, "nobh"}, |
1276 | {Opt_bh, "bh"}, | |
1277 | {Opt_commit, "commit=%u"}, | |
30773840 TT |
1278 | {Opt_min_batch_time, "min_batch_time=%u"}, |
1279 | {Opt_max_batch_time, "max_batch_time=%u"}, | |
ac27a0ec | 1280 | {Opt_journal_update, "journal=update"}, |
ac27a0ec | 1281 | {Opt_journal_dev, "journal_dev=%u"}, |
818d276c GS |
1282 | {Opt_journal_checksum, "journal_checksum"}, |
1283 | {Opt_journal_async_commit, "journal_async_commit"}, | |
ac27a0ec DK |
1284 | {Opt_abort, "abort"}, |
1285 | {Opt_data_journal, "data=journal"}, | |
1286 | {Opt_data_ordered, "data=ordered"}, | |
1287 | {Opt_data_writeback, "data=writeback"}, | |
5bf5683a HK |
1288 | {Opt_data_err_abort, "data_err=abort"}, |
1289 | {Opt_data_err_ignore, "data_err=ignore"}, | |
ac27a0ec DK |
1290 | {Opt_offusrjquota, "usrjquota="}, |
1291 | {Opt_usrjquota, "usrjquota=%s"}, | |
1292 | {Opt_offgrpjquota, "grpjquota="}, | |
1293 | {Opt_grpjquota, "grpjquota=%s"}, | |
1294 | {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, | |
1295 | {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, | |
5a20bdfc | 1296 | {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, |
ac27a0ec DK |
1297 | {Opt_grpquota, "grpquota"}, |
1298 | {Opt_noquota, "noquota"}, | |
1299 | {Opt_quota, "quota"}, | |
1300 | {Opt_usrquota, "usrquota"}, | |
1301 | {Opt_barrier, "barrier=%u"}, | |
06705bff TT |
1302 | {Opt_barrier, "barrier"}, |
1303 | {Opt_nobarrier, "nobarrier"}, | |
25ec56b5 | 1304 | {Opt_i_version, "i_version"}, |
c9de560d | 1305 | {Opt_stripe, "stripe=%u"}, |
ac27a0ec | 1306 | {Opt_resize, "resize"}, |
64769240 | 1307 | {Opt_delalloc, "delalloc"}, |
dd919b98 | 1308 | {Opt_nodelalloc, "nodelalloc"}, |
1449032b TT |
1309 | {Opt_mblk_io_submit, "mblk_io_submit"}, |
1310 | {Opt_nomblk_io_submit, "nomblk_io_submit"}, | |
6fd058f7 TT |
1311 | {Opt_block_validity, "block_validity"}, |
1312 | {Opt_noblock_validity, "noblock_validity"}, | |
240799cd | 1313 | {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, |
b3881f74 | 1314 | {Opt_journal_ioprio, "journal_ioprio=%u"}, |
afd4672d | 1315 | {Opt_auto_da_alloc, "auto_da_alloc=%u"}, |
06705bff TT |
1316 | {Opt_auto_da_alloc, "auto_da_alloc"}, |
1317 | {Opt_noauto_da_alloc, "noauto_da_alloc"}, | |
744692dc JZ |
1318 | {Opt_dioread_nolock, "dioread_nolock"}, |
1319 | {Opt_dioread_lock, "dioread_lock"}, | |
5328e635 ES |
1320 | {Opt_discard, "discard"}, |
1321 | {Opt_nodiscard, "nodiscard"}, | |
bfff6873 LC |
1322 | {Opt_init_inode_table, "init_itable=%u"}, |
1323 | {Opt_init_inode_table, "init_itable"}, | |
1324 | {Opt_noinit_inode_table, "noinit_itable"}, | |
f3f12faa | 1325 | {Opt_err, NULL}, |
ac27a0ec DK |
1326 | }; |
1327 | ||
617ba13b | 1328 | static ext4_fsblk_t get_sb_block(void **data) |
ac27a0ec | 1329 | { |
617ba13b | 1330 | ext4_fsblk_t sb_block; |
ac27a0ec DK |
1331 | char *options = (char *) *data; |
1332 | ||
1333 | if (!options || strncmp(options, "sb=", 3) != 0) | |
1334 | return 1; /* Default location */ | |
0b8e58a1 | 1335 | |
ac27a0ec | 1336 | options += 3; |
0b8e58a1 | 1337 | /* TODO: use simple_strtoll with >32bit ext4 */ |
ac27a0ec DK |
1338 | sb_block = simple_strtoul(options, &options, 0); |
1339 | if (*options && *options != ',') { | |
4776004f | 1340 | printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", |
ac27a0ec DK |
1341 | (char *) *data); |
1342 | return 1; | |
1343 | } | |
1344 | if (*options == ',') | |
1345 | options++; | |
1346 | *data = (void *) options; | |
0b8e58a1 | 1347 | |
ac27a0ec DK |
1348 | return sb_block; |
1349 | } | |
1350 | ||
b3881f74 | 1351 | #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) |
437ca0fd DM |
1352 | static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n" |
1353 | "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; | |
b3881f74 | 1354 | |
56c50f11 DM |
1355 | #ifdef CONFIG_QUOTA |
1356 | static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) | |
1357 | { | |
1358 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1359 | char *qname; | |
1360 | ||
1361 | if (sb_any_quota_loaded(sb) && | |
1362 | !sbi->s_qf_names[qtype]) { | |
1363 | ext4_msg(sb, KERN_ERR, | |
1364 | "Cannot change journaled " | |
1365 | "quota options when quota turned on"); | |
1366 | return 0; | |
1367 | } | |
1368 | qname = match_strdup(args); | |
1369 | if (!qname) { | |
1370 | ext4_msg(sb, KERN_ERR, | |
1371 | "Not enough memory for storing quotafile name"); | |
1372 | return 0; | |
1373 | } | |
1374 | if (sbi->s_qf_names[qtype] && | |
1375 | strcmp(sbi->s_qf_names[qtype], qname)) { | |
1376 | ext4_msg(sb, KERN_ERR, | |
1377 | "%s quota file already specified", QTYPE2NAME(qtype)); | |
1378 | kfree(qname); | |
1379 | return 0; | |
1380 | } | |
1381 | sbi->s_qf_names[qtype] = qname; | |
1382 | if (strchr(sbi->s_qf_names[qtype], '/')) { | |
1383 | ext4_msg(sb, KERN_ERR, | |
1384 | "quotafile must be on filesystem root"); | |
1385 | kfree(sbi->s_qf_names[qtype]); | |
1386 | sbi->s_qf_names[qtype] = NULL; | |
1387 | return 0; | |
1388 | } | |
1389 | set_opt(sbi->s_mount_opt, QUOTA); | |
1390 | return 1; | |
1391 | } | |
1392 | ||
1393 | static int clear_qf_name(struct super_block *sb, int qtype) | |
1394 | { | |
1395 | ||
1396 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1397 | ||
1398 | if (sb_any_quota_loaded(sb) && | |
1399 | sbi->s_qf_names[qtype]) { | |
1400 | ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" | |
1401 | " when quota turned on"); | |
1402 | return 0; | |
1403 | } | |
1404 | /* | |
1405 | * The space will be released later when all options are confirmed | |
1406 | * to be correct | |
1407 | */ | |
1408 | sbi->s_qf_names[qtype] = NULL; | |
1409 | return 1; | |
1410 | } | |
1411 | #endif | |
1412 | ||
2b2d6d01 | 1413 | static int parse_options(char *options, struct super_block *sb, |
c3191067 | 1414 | unsigned long *journal_devnum, |
b3881f74 | 1415 | unsigned int *journal_ioprio, |
2b2d6d01 | 1416 | ext4_fsblk_t *n_blocks_count, int is_remount) |
ac27a0ec | 1417 | { |
617ba13b | 1418 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
2b2d6d01 | 1419 | char *p; |
ac27a0ec DK |
1420 | substring_t args[MAX_OPT_ARGS]; |
1421 | int data_opt = 0; | |
1422 | int option; | |
1423 | #ifdef CONFIG_QUOTA | |
56c50f11 | 1424 | int qfmt; |
ac27a0ec DK |
1425 | #endif |
1426 | ||
1427 | if (!options) | |
1428 | return 1; | |
1429 | ||
2b2d6d01 | 1430 | while ((p = strsep(&options, ",")) != NULL) { |
ac27a0ec DK |
1431 | int token; |
1432 | if (!*p) | |
1433 | continue; | |
1434 | ||
15121c18 ES |
1435 | /* |
1436 | * Initialize args struct so we know whether arg was | |
1437 | * found; some options take optional arguments. | |
1438 | */ | |
1439 | args[0].to = args[0].from = 0; | |
ac27a0ec DK |
1440 | token = match_token(p, tokens, args); |
1441 | switch (token) { | |
1442 | case Opt_bsd_df: | |
437ca0fd | 1443 | ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); |
2b2d6d01 | 1444 | clear_opt(sbi->s_mount_opt, MINIX_DF); |
ac27a0ec DK |
1445 | break; |
1446 | case Opt_minix_df: | |
437ca0fd | 1447 | ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); |
2b2d6d01 | 1448 | set_opt(sbi->s_mount_opt, MINIX_DF); |
437ca0fd | 1449 | |
ac27a0ec DK |
1450 | break; |
1451 | case Opt_grpid: | |
437ca0fd | 1452 | ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); |
2b2d6d01 | 1453 | set_opt(sbi->s_mount_opt, GRPID); |
437ca0fd | 1454 | |
ac27a0ec DK |
1455 | break; |
1456 | case Opt_nogrpid: | |
437ca0fd | 1457 | ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); |
2b2d6d01 | 1458 | clear_opt(sbi->s_mount_opt, GRPID); |
437ca0fd | 1459 | |
ac27a0ec DK |
1460 | break; |
1461 | case Opt_resuid: | |
1462 | if (match_int(&args[0], &option)) | |
1463 | return 0; | |
1464 | sbi->s_resuid = option; | |
1465 | break; | |
1466 | case Opt_resgid: | |
1467 | if (match_int(&args[0], &option)) | |
1468 | return 0; | |
1469 | sbi->s_resgid = option; | |
1470 | break; | |
1471 | case Opt_sb: | |
1472 | /* handled by get_sb_block() instead of here */ | |
1473 | /* *sb_block = match_int(&args[0]); */ | |
1474 | break; | |
1475 | case Opt_err_panic: | |
2b2d6d01 TT |
1476 | clear_opt(sbi->s_mount_opt, ERRORS_CONT); |
1477 | clear_opt(sbi->s_mount_opt, ERRORS_RO); | |
1478 | set_opt(sbi->s_mount_opt, ERRORS_PANIC); | |
ac27a0ec DK |
1479 | break; |
1480 | case Opt_err_ro: | |
2b2d6d01 TT |
1481 | clear_opt(sbi->s_mount_opt, ERRORS_CONT); |
1482 | clear_opt(sbi->s_mount_opt, ERRORS_PANIC); | |
1483 | set_opt(sbi->s_mount_opt, ERRORS_RO); | |
ac27a0ec DK |
1484 | break; |
1485 | case Opt_err_cont: | |
2b2d6d01 TT |
1486 | clear_opt(sbi->s_mount_opt, ERRORS_RO); |
1487 | clear_opt(sbi->s_mount_opt, ERRORS_PANIC); | |
1488 | set_opt(sbi->s_mount_opt, ERRORS_CONT); | |
ac27a0ec DK |
1489 | break; |
1490 | case Opt_nouid32: | |
2b2d6d01 | 1491 | set_opt(sbi->s_mount_opt, NO_UID32); |
ac27a0ec | 1492 | break; |
ac27a0ec | 1493 | case Opt_debug: |
2b2d6d01 | 1494 | set_opt(sbi->s_mount_opt, DEBUG); |
ac27a0ec DK |
1495 | break; |
1496 | case Opt_oldalloc: | |
2b2d6d01 | 1497 | set_opt(sbi->s_mount_opt, OLDALLOC); |
ac27a0ec DK |
1498 | break; |
1499 | case Opt_orlov: | |
2b2d6d01 | 1500 | clear_opt(sbi->s_mount_opt, OLDALLOC); |
ac27a0ec | 1501 | break; |
03010a33 | 1502 | #ifdef CONFIG_EXT4_FS_XATTR |
ac27a0ec | 1503 | case Opt_user_xattr: |
2b2d6d01 | 1504 | set_opt(sbi->s_mount_opt, XATTR_USER); |
ac27a0ec DK |
1505 | break; |
1506 | case Opt_nouser_xattr: | |
2b2d6d01 | 1507 | clear_opt(sbi->s_mount_opt, XATTR_USER); |
ac27a0ec DK |
1508 | break; |
1509 | #else | |
1510 | case Opt_user_xattr: | |
1511 | case Opt_nouser_xattr: | |
b31e1552 | 1512 | ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported"); |
ac27a0ec DK |
1513 | break; |
1514 | #endif | |
03010a33 | 1515 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
ac27a0ec DK |
1516 | case Opt_acl: |
1517 | set_opt(sbi->s_mount_opt, POSIX_ACL); | |
1518 | break; | |
1519 | case Opt_noacl: | |
1520 | clear_opt(sbi->s_mount_opt, POSIX_ACL); | |
1521 | break; | |
1522 | #else | |
1523 | case Opt_acl: | |
1524 | case Opt_noacl: | |
b31e1552 | 1525 | ext4_msg(sb, KERN_ERR, "(no)acl options not supported"); |
ac27a0ec DK |
1526 | break; |
1527 | #endif | |
ac27a0ec DK |
1528 | case Opt_journal_update: |
1529 | /* @@@ FIXME */ | |
1530 | /* Eventually we will want to be able to create | |
1531 | a journal file here. For now, only allow the | |
1532 | user to specify an existing inode to be the | |
1533 | journal file. */ | |
1534 | if (is_remount) { | |
b31e1552 ES |
1535 | ext4_msg(sb, KERN_ERR, |
1536 | "Cannot specify journal on remount"); | |
ac27a0ec DK |
1537 | return 0; |
1538 | } | |
2b2d6d01 | 1539 | set_opt(sbi->s_mount_opt, UPDATE_JOURNAL); |
ac27a0ec | 1540 | break; |
ac27a0ec DK |
1541 | case Opt_journal_dev: |
1542 | if (is_remount) { | |
b31e1552 ES |
1543 | ext4_msg(sb, KERN_ERR, |
1544 | "Cannot specify journal on remount"); | |
ac27a0ec DK |
1545 | return 0; |
1546 | } | |
1547 | if (match_int(&args[0], &option)) | |
1548 | return 0; | |
1549 | *journal_devnum = option; | |
1550 | break; | |
818d276c | 1551 | case Opt_journal_checksum: |
d4da6c9c LT |
1552 | set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM); |
1553 | break; | |
818d276c GS |
1554 | case Opt_journal_async_commit: |
1555 | set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT); | |
d4da6c9c | 1556 | set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM); |
818d276c | 1557 | break; |
ac27a0ec | 1558 | case Opt_noload: |
2b2d6d01 | 1559 | set_opt(sbi->s_mount_opt, NOLOAD); |
ac27a0ec DK |
1560 | break; |
1561 | case Opt_commit: | |
1562 | if (match_int(&args[0], &option)) | |
1563 | return 0; | |
1564 | if (option < 0) | |
1565 | return 0; | |
1566 | if (option == 0) | |
cd02ff0b | 1567 | option = JBD2_DEFAULT_MAX_COMMIT_AGE; |
ac27a0ec DK |
1568 | sbi->s_commit_interval = HZ * option; |
1569 | break; | |
30773840 TT |
1570 | case Opt_max_batch_time: |
1571 | if (match_int(&args[0], &option)) | |
1572 | return 0; | |
1573 | if (option < 0) | |
1574 | return 0; | |
1575 | if (option == 0) | |
1576 | option = EXT4_DEF_MAX_BATCH_TIME; | |
1577 | sbi->s_max_batch_time = option; | |
1578 | break; | |
1579 | case Opt_min_batch_time: | |
1580 | if (match_int(&args[0], &option)) | |
1581 | return 0; | |
1582 | if (option < 0) | |
1583 | return 0; | |
1584 | sbi->s_min_batch_time = option; | |
1585 | break; | |
ac27a0ec | 1586 | case Opt_data_journal: |
617ba13b | 1587 | data_opt = EXT4_MOUNT_JOURNAL_DATA; |
ac27a0ec DK |
1588 | goto datacheck; |
1589 | case Opt_data_ordered: | |
617ba13b | 1590 | data_opt = EXT4_MOUNT_ORDERED_DATA; |
ac27a0ec DK |
1591 | goto datacheck; |
1592 | case Opt_data_writeback: | |
617ba13b | 1593 | data_opt = EXT4_MOUNT_WRITEBACK_DATA; |
ac27a0ec DK |
1594 | datacheck: |
1595 | if (is_remount) { | |
482a7425 | 1596 | if (test_opt(sb, DATA_FLAGS) != data_opt) { |
b31e1552 ES |
1597 | ext4_msg(sb, KERN_ERR, |
1598 | "Cannot change data mode on remount"); | |
ac27a0ec DK |
1599 | return 0; |
1600 | } | |
1601 | } else { | |
482a7425 | 1602 | clear_opt(sbi->s_mount_opt, DATA_FLAGS); |
ac27a0ec DK |
1603 | sbi->s_mount_opt |= data_opt; |
1604 | } | |
1605 | break; | |
5bf5683a HK |
1606 | case Opt_data_err_abort: |
1607 | set_opt(sbi->s_mount_opt, DATA_ERR_ABORT); | |
1608 | break; | |
1609 | case Opt_data_err_ignore: | |
1610 | clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT); | |
1611 | break; | |
ac27a0ec DK |
1612 | #ifdef CONFIG_QUOTA |
1613 | case Opt_usrjquota: | |
56c50f11 | 1614 | if (!set_qf_name(sb, USRQUOTA, &args[0])) |
ac27a0ec | 1615 | return 0; |
56c50f11 DM |
1616 | break; |
1617 | case Opt_grpjquota: | |
1618 | if (!set_qf_name(sb, GRPQUOTA, &args[0])) | |
ac27a0ec | 1619 | return 0; |
ac27a0ec DK |
1620 | break; |
1621 | case Opt_offusrjquota: | |
56c50f11 DM |
1622 | if (!clear_qf_name(sb, USRQUOTA)) |
1623 | return 0; | |
1624 | break; | |
ac27a0ec | 1625 | case Opt_offgrpjquota: |
56c50f11 | 1626 | if (!clear_qf_name(sb, GRPQUOTA)) |
ac27a0ec | 1627 | return 0; |
ac27a0ec | 1628 | break; |
56c50f11 | 1629 | |
ac27a0ec | 1630 | case Opt_jqfmt_vfsold: |
dfc5d03f JK |
1631 | qfmt = QFMT_VFS_OLD; |
1632 | goto set_qf_format; | |
ac27a0ec | 1633 | case Opt_jqfmt_vfsv0: |
dfc5d03f | 1634 | qfmt = QFMT_VFS_V0; |
5a20bdfc JK |
1635 | goto set_qf_format; |
1636 | case Opt_jqfmt_vfsv1: | |
1637 | qfmt = QFMT_VFS_V1; | |
dfc5d03f | 1638 | set_qf_format: |
17bd13b3 | 1639 | if (sb_any_quota_loaded(sb) && |
dfc5d03f | 1640 | sbi->s_jquota_fmt != qfmt) { |
b31e1552 | 1641 | ext4_msg(sb, KERN_ERR, "Cannot change " |
dfc5d03f | 1642 | "journaled quota options when " |
b31e1552 | 1643 | "quota turned on"); |
dfc5d03f JK |
1644 | return 0; |
1645 | } | |
1646 | sbi->s_jquota_fmt = qfmt; | |
ac27a0ec DK |
1647 | break; |
1648 | case Opt_quota: | |
1649 | case Opt_usrquota: | |
1650 | set_opt(sbi->s_mount_opt, QUOTA); | |
1651 | set_opt(sbi->s_mount_opt, USRQUOTA); | |
1652 | break; | |
1653 | case Opt_grpquota: | |
1654 | set_opt(sbi->s_mount_opt, QUOTA); | |
1655 | set_opt(sbi->s_mount_opt, GRPQUOTA); | |
1656 | break; | |
1657 | case Opt_noquota: | |
17bd13b3 | 1658 | if (sb_any_quota_loaded(sb)) { |
b31e1552 ES |
1659 | ext4_msg(sb, KERN_ERR, "Cannot change quota " |
1660 | "options when quota turned on"); | |
ac27a0ec DK |
1661 | return 0; |
1662 | } | |
1663 | clear_opt(sbi->s_mount_opt, QUOTA); | |
1664 | clear_opt(sbi->s_mount_opt, USRQUOTA); | |
1665 | clear_opt(sbi->s_mount_opt, GRPQUOTA); | |
1666 | break; | |
1667 | #else | |
1668 | case Opt_quota: | |
1669 | case Opt_usrquota: | |
1670 | case Opt_grpquota: | |
b31e1552 ES |
1671 | ext4_msg(sb, KERN_ERR, |
1672 | "quota options not supported"); | |
cd59e7b9 | 1673 | break; |
ac27a0ec DK |
1674 | case Opt_usrjquota: |
1675 | case Opt_grpjquota: | |
1676 | case Opt_offusrjquota: | |
1677 | case Opt_offgrpjquota: | |
1678 | case Opt_jqfmt_vfsold: | |
1679 | case Opt_jqfmt_vfsv0: | |
5a20bdfc | 1680 | case Opt_jqfmt_vfsv1: |
b31e1552 ES |
1681 | ext4_msg(sb, KERN_ERR, |
1682 | "journaled quota options not supported"); | |
ac27a0ec DK |
1683 | break; |
1684 | case Opt_noquota: | |
1685 | break; | |
1686 | #endif | |
1687 | case Opt_abort: | |
4ab2f15b | 1688 | sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; |
ac27a0ec | 1689 | break; |
06705bff TT |
1690 | case Opt_nobarrier: |
1691 | clear_opt(sbi->s_mount_opt, BARRIER); | |
1692 | break; | |
ac27a0ec | 1693 | case Opt_barrier: |
15121c18 ES |
1694 | if (args[0].from) { |
1695 | if (match_int(&args[0], &option)) | |
1696 | return 0; | |
1697 | } else | |
1698 | option = 1; /* No argument, default to 1 */ | |
ac27a0ec DK |
1699 | if (option) |
1700 | set_opt(sbi->s_mount_opt, BARRIER); | |
1701 | else | |
1702 | clear_opt(sbi->s_mount_opt, BARRIER); | |
1703 | break; | |
1704 | case Opt_ignore: | |
1705 | break; | |
1706 | case Opt_resize: | |
1707 | if (!is_remount) { | |
b31e1552 ES |
1708 | ext4_msg(sb, KERN_ERR, |
1709 | "resize option only available " | |
1710 | "for remount"); | |
ac27a0ec DK |
1711 | return 0; |
1712 | } | |
1713 | if (match_int(&args[0], &option) != 0) | |
1714 | return 0; | |
1715 | *n_blocks_count = option; | |
1716 | break; | |
1717 | case Opt_nobh: | |
206f7ab4 CH |
1718 | ext4_msg(sb, KERN_WARNING, |
1719 | "Ignoring deprecated nobh option"); | |
ac27a0ec DK |
1720 | break; |
1721 | case Opt_bh: | |
206f7ab4 CH |
1722 | ext4_msg(sb, KERN_WARNING, |
1723 | "Ignoring deprecated bh option"); | |
ac27a0ec | 1724 | break; |
25ec56b5 JNC |
1725 | case Opt_i_version: |
1726 | set_opt(sbi->s_mount_opt, I_VERSION); | |
1727 | sb->s_flags |= MS_I_VERSION; | |
1728 | break; | |
dd919b98 AK |
1729 | case Opt_nodelalloc: |
1730 | clear_opt(sbi->s_mount_opt, DELALLOC); | |
1731 | break; | |
1449032b TT |
1732 | case Opt_mblk_io_submit: |
1733 | set_opt(sbi->s_mount_opt, MBLK_IO_SUBMIT); | |
1734 | break; | |
1735 | case Opt_nomblk_io_submit: | |
1736 | clear_opt(sbi->s_mount_opt, MBLK_IO_SUBMIT); | |
1737 | break; | |
c9de560d AT |
1738 | case Opt_stripe: |
1739 | if (match_int(&args[0], &option)) | |
1740 | return 0; | |
1741 | if (option < 0) | |
1742 | return 0; | |
1743 | sbi->s_stripe = option; | |
1744 | break; | |
64769240 AT |
1745 | case Opt_delalloc: |
1746 | set_opt(sbi->s_mount_opt, DELALLOC); | |
1747 | break; | |
6fd058f7 TT |
1748 | case Opt_block_validity: |
1749 | set_opt(sbi->s_mount_opt, BLOCK_VALIDITY); | |
1750 | break; | |
1751 | case Opt_noblock_validity: | |
1752 | clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY); | |
1753 | break; | |
240799cd TT |
1754 | case Opt_inode_readahead_blks: |
1755 | if (match_int(&args[0], &option)) | |
1756 | return 0; | |
1757 | if (option < 0 || option > (1 << 30)) | |
1758 | return 0; | |
f7c43950 | 1759 | if (!is_power_of_2(option)) { |
b31e1552 ES |
1760 | ext4_msg(sb, KERN_ERR, |
1761 | "EXT4-fs: inode_readahead_blks" | |
1762 | " must be a power of 2"); | |
3197ebdb TT |
1763 | return 0; |
1764 | } | |
240799cd TT |
1765 | sbi->s_inode_readahead_blks = option; |
1766 | break; | |
b3881f74 TT |
1767 | case Opt_journal_ioprio: |
1768 | if (match_int(&args[0], &option)) | |
1769 | return 0; | |
1770 | if (option < 0 || option > 7) | |
1771 | break; | |
1772 | *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, | |
1773 | option); | |
1774 | break; | |
06705bff TT |
1775 | case Opt_noauto_da_alloc: |
1776 | set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC); | |
1777 | break; | |
afd4672d | 1778 | case Opt_auto_da_alloc: |
15121c18 ES |
1779 | if (args[0].from) { |
1780 | if (match_int(&args[0], &option)) | |
1781 | return 0; | |
1782 | } else | |
1783 | option = 1; /* No argument, default to 1 */ | |
afd4672d TT |
1784 | if (option) |
1785 | clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC); | |
1786 | else | |
1787 | set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC); | |
1788 | break; | |
5328e635 ES |
1789 | case Opt_discard: |
1790 | set_opt(sbi->s_mount_opt, DISCARD); | |
1791 | break; | |
1792 | case Opt_nodiscard: | |
1793 | clear_opt(sbi->s_mount_opt, DISCARD); | |
1794 | break; | |
744692dc JZ |
1795 | case Opt_dioread_nolock: |
1796 | set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK); | |
1797 | break; | |
1798 | case Opt_dioread_lock: | |
1799 | clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK); | |
1800 | break; | |
bfff6873 LC |
1801 | case Opt_init_inode_table: |
1802 | set_opt(sbi->s_mount_opt, INIT_INODE_TABLE); | |
1803 | if (args[0].from) { | |
1804 | if (match_int(&args[0], &option)) | |
1805 | return 0; | |
1806 | } else | |
1807 | option = EXT4_DEF_LI_WAIT_MULT; | |
1808 | if (option < 0) | |
1809 | return 0; | |
1810 | sbi->s_li_wait_mult = option; | |
1811 | break; | |
1812 | case Opt_noinit_inode_table: | |
1813 | clear_opt(sbi->s_mount_opt, INIT_INODE_TABLE); | |
1814 | break; | |
ac27a0ec | 1815 | default: |
b31e1552 ES |
1816 | ext4_msg(sb, KERN_ERR, |
1817 | "Unrecognized mount option \"%s\" " | |
1818 | "or missing value", p); | |
ac27a0ec DK |
1819 | return 0; |
1820 | } | |
1821 | } | |
1822 | #ifdef CONFIG_QUOTA | |
1823 | if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { | |
482a7425 | 1824 | if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA]) |
ac27a0ec DK |
1825 | clear_opt(sbi->s_mount_opt, USRQUOTA); |
1826 | ||
482a7425 | 1827 | if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA]) |
ac27a0ec DK |
1828 | clear_opt(sbi->s_mount_opt, GRPQUOTA); |
1829 | ||
56c50f11 | 1830 | if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { |
b31e1552 ES |
1831 | ext4_msg(sb, KERN_ERR, "old and new quota " |
1832 | "format mixing"); | |
ac27a0ec DK |
1833 | return 0; |
1834 | } | |
1835 | ||
1836 | if (!sbi->s_jquota_fmt) { | |
b31e1552 ES |
1837 | ext4_msg(sb, KERN_ERR, "journaled quota format " |
1838 | "not specified"); | |
ac27a0ec DK |
1839 | return 0; |
1840 | } | |
1841 | } else { | |
1842 | if (sbi->s_jquota_fmt) { | |
b31e1552 | 1843 | ext4_msg(sb, KERN_ERR, "journaled quota format " |
2c8be6b2 | 1844 | "specified with no journaling " |
b31e1552 | 1845 | "enabled"); |
ac27a0ec DK |
1846 | return 0; |
1847 | } | |
1848 | } | |
1849 | #endif | |
1850 | return 1; | |
1851 | } | |
1852 | ||
617ba13b | 1853 | static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, |
ac27a0ec DK |
1854 | int read_only) |
1855 | { | |
617ba13b | 1856 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec DK |
1857 | int res = 0; |
1858 | ||
617ba13b | 1859 | if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { |
b31e1552 ES |
1860 | ext4_msg(sb, KERN_ERR, "revision level too high, " |
1861 | "forcing read-only mode"); | |
ac27a0ec DK |
1862 | res = MS_RDONLY; |
1863 | } | |
1864 | if (read_only) | |
1865 | return res; | |
617ba13b | 1866 | if (!(sbi->s_mount_state & EXT4_VALID_FS)) |
b31e1552 ES |
1867 | ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " |
1868 | "running e2fsck is recommended"); | |
617ba13b | 1869 | else if ((sbi->s_mount_state & EXT4_ERROR_FS)) |
b31e1552 ES |
1870 | ext4_msg(sb, KERN_WARNING, |
1871 | "warning: mounting fs with errors, " | |
1872 | "running e2fsck is recommended"); | |
ac27a0ec DK |
1873 | else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 && |
1874 | le16_to_cpu(es->s_mnt_count) >= | |
1875 | (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) | |
b31e1552 ES |
1876 | ext4_msg(sb, KERN_WARNING, |
1877 | "warning: maximal mount count reached, " | |
1878 | "running e2fsck is recommended"); | |
ac27a0ec DK |
1879 | else if (le32_to_cpu(es->s_checkinterval) && |
1880 | (le32_to_cpu(es->s_lastcheck) + | |
1881 | le32_to_cpu(es->s_checkinterval) <= get_seconds())) | |
b31e1552 ES |
1882 | ext4_msg(sb, KERN_WARNING, |
1883 | "warning: checktime reached, " | |
1884 | "running e2fsck is recommended"); | |
0b8e58a1 | 1885 | if (!sbi->s_journal) |
0390131b | 1886 | es->s_state &= cpu_to_le16(~EXT4_VALID_FS); |
ac27a0ec | 1887 | if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) |
617ba13b | 1888 | es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); |
e8546d06 | 1889 | le16_add_cpu(&es->s_mnt_count, 1); |
ac27a0ec | 1890 | es->s_mtime = cpu_to_le32(get_seconds()); |
617ba13b | 1891 | ext4_update_dynamic_rev(sb); |
0390131b FM |
1892 | if (sbi->s_journal) |
1893 | EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
ac27a0ec | 1894 | |
e2d67052 | 1895 | ext4_commit_super(sb, 1); |
ac27a0ec | 1896 | if (test_opt(sb, DEBUG)) |
a9df9a49 | 1897 | printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " |
7f4520cc | 1898 | "bpg=%lu, ipg=%lu, mo=%04x]\n", |
ac27a0ec DK |
1899 | sb->s_blocksize, |
1900 | sbi->s_groups_count, | |
617ba13b MC |
1901 | EXT4_BLOCKS_PER_GROUP(sb), |
1902 | EXT4_INODES_PER_GROUP(sb), | |
ac27a0ec DK |
1903 | sbi->s_mount_opt); |
1904 | ||
ac27a0ec DK |
1905 | return res; |
1906 | } | |
1907 | ||
772cb7c8 JS |
1908 | static int ext4_fill_flex_info(struct super_block *sb) |
1909 | { | |
1910 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1911 | struct ext4_group_desc *gdp = NULL; | |
772cb7c8 JS |
1912 | ext4_group_t flex_group_count; |
1913 | ext4_group_t flex_group; | |
1914 | int groups_per_flex = 0; | |
c5ca7c76 | 1915 | size_t size; |
772cb7c8 JS |
1916 | int i; |
1917 | ||
503358ae TT |
1918 | sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; |
1919 | groups_per_flex = 1 << sbi->s_log_groups_per_flex; | |
1920 | ||
1921 | if (groups_per_flex < 2) { | |
772cb7c8 JS |
1922 | sbi->s_log_groups_per_flex = 0; |
1923 | return 1; | |
1924 | } | |
1925 | ||
c62a11fd FB |
1926 | /* We allocate both existing and potentially added groups */ |
1927 | flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) + | |
d94e99a6 AK |
1928 | ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) << |
1929 | EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex; | |
c5ca7c76 TT |
1930 | size = flex_group_count * sizeof(struct flex_groups); |
1931 | sbi->s_flex_groups = kzalloc(size, GFP_KERNEL); | |
1932 | if (sbi->s_flex_groups == NULL) { | |
1933 | sbi->s_flex_groups = vmalloc(size); | |
1934 | if (sbi->s_flex_groups) | |
1935 | memset(sbi->s_flex_groups, 0, size); | |
1936 | } | |
772cb7c8 | 1937 | if (sbi->s_flex_groups == NULL) { |
b31e1552 ES |
1938 | ext4_msg(sb, KERN_ERR, "not enough memory for " |
1939 | "%u flex groups", flex_group_count); | |
772cb7c8 JS |
1940 | goto failed; |
1941 | } | |
772cb7c8 | 1942 | |
772cb7c8 | 1943 | for (i = 0; i < sbi->s_groups_count; i++) { |
88b6edd1 | 1944 | gdp = ext4_get_group_desc(sb, i, NULL); |
772cb7c8 JS |
1945 | |
1946 | flex_group = ext4_flex_group(sbi, i); | |
7ad9bb65 TT |
1947 | atomic_add(ext4_free_inodes_count(sb, gdp), |
1948 | &sbi->s_flex_groups[flex_group].free_inodes); | |
1949 | atomic_add(ext4_free_blks_count(sb, gdp), | |
1950 | &sbi->s_flex_groups[flex_group].free_blocks); | |
1951 | atomic_add(ext4_used_dirs_count(sb, gdp), | |
1952 | &sbi->s_flex_groups[flex_group].used_dirs); | |
772cb7c8 JS |
1953 | } |
1954 | ||
1955 | return 1; | |
1956 | failed: | |
1957 | return 0; | |
1958 | } | |
1959 | ||
717d50e4 AD |
1960 | __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group, |
1961 | struct ext4_group_desc *gdp) | |
1962 | { | |
1963 | __u16 crc = 0; | |
1964 | ||
1965 | if (sbi->s_es->s_feature_ro_compat & | |
1966 | cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) { | |
1967 | int offset = offsetof(struct ext4_group_desc, bg_checksum); | |
1968 | __le32 le_group = cpu_to_le32(block_group); | |
1969 | ||
1970 | crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); | |
1971 | crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); | |
1972 | crc = crc16(crc, (__u8 *)gdp, offset); | |
1973 | offset += sizeof(gdp->bg_checksum); /* skip checksum */ | |
1974 | /* for checksum of struct ext4_group_desc do the rest...*/ | |
1975 | if ((sbi->s_es->s_feature_incompat & | |
1976 | cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) && | |
1977 | offset < le16_to_cpu(sbi->s_es->s_desc_size)) | |
1978 | crc = crc16(crc, (__u8 *)gdp + offset, | |
1979 | le16_to_cpu(sbi->s_es->s_desc_size) - | |
1980 | offset); | |
1981 | } | |
1982 | ||
1983 | return cpu_to_le16(crc); | |
1984 | } | |
1985 | ||
1986 | int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group, | |
1987 | struct ext4_group_desc *gdp) | |
1988 | { | |
1989 | if ((sbi->s_es->s_feature_ro_compat & | |
1990 | cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) && | |
1991 | (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp))) | |
1992 | return 0; | |
1993 | ||
1994 | return 1; | |
1995 | } | |
1996 | ||
ac27a0ec | 1997 | /* Called at mount-time, super-block is locked */ |
bfff6873 LC |
1998 | static int ext4_check_descriptors(struct super_block *sb, |
1999 | ext4_group_t *first_not_zeroed) | |
ac27a0ec | 2000 | { |
617ba13b MC |
2001 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
2002 | ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); | |
2003 | ext4_fsblk_t last_block; | |
bd81d8ee LV |
2004 | ext4_fsblk_t block_bitmap; |
2005 | ext4_fsblk_t inode_bitmap; | |
2006 | ext4_fsblk_t inode_table; | |
ce421581 | 2007 | int flexbg_flag = 0; |
bfff6873 | 2008 | ext4_group_t i, grp = sbi->s_groups_count; |
ac27a0ec | 2009 | |
ce421581 JS |
2010 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) |
2011 | flexbg_flag = 1; | |
2012 | ||
af5bc92d | 2013 | ext4_debug("Checking group descriptors"); |
ac27a0ec | 2014 | |
197cd65a AM |
2015 | for (i = 0; i < sbi->s_groups_count; i++) { |
2016 | struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); | |
2017 | ||
ce421581 | 2018 | if (i == sbi->s_groups_count - 1 || flexbg_flag) |
bd81d8ee | 2019 | last_block = ext4_blocks_count(sbi->s_es) - 1; |
ac27a0ec DK |
2020 | else |
2021 | last_block = first_block + | |
617ba13b | 2022 | (EXT4_BLOCKS_PER_GROUP(sb) - 1); |
ac27a0ec | 2023 | |
bfff6873 LC |
2024 | if ((grp == sbi->s_groups_count) && |
2025 | !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) | |
2026 | grp = i; | |
2027 | ||
8fadc143 | 2028 | block_bitmap = ext4_block_bitmap(sb, gdp); |
2b2d6d01 | 2029 | if (block_bitmap < first_block || block_bitmap > last_block) { |
b31e1552 | 2030 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 2031 | "Block bitmap for group %u not in group " |
b31e1552 | 2032 | "(block %llu)!", i, block_bitmap); |
ac27a0ec DK |
2033 | return 0; |
2034 | } | |
8fadc143 | 2035 | inode_bitmap = ext4_inode_bitmap(sb, gdp); |
2b2d6d01 | 2036 | if (inode_bitmap < first_block || inode_bitmap > last_block) { |
b31e1552 | 2037 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 2038 | "Inode bitmap for group %u not in group " |
b31e1552 | 2039 | "(block %llu)!", i, inode_bitmap); |
ac27a0ec DK |
2040 | return 0; |
2041 | } | |
8fadc143 | 2042 | inode_table = ext4_inode_table(sb, gdp); |
bd81d8ee | 2043 | if (inode_table < first_block || |
2b2d6d01 | 2044 | inode_table + sbi->s_itb_per_group - 1 > last_block) { |
b31e1552 | 2045 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 2046 | "Inode table for group %u not in group " |
b31e1552 | 2047 | "(block %llu)!", i, inode_table); |
ac27a0ec DK |
2048 | return 0; |
2049 | } | |
955ce5f5 | 2050 | ext4_lock_group(sb, i); |
717d50e4 | 2051 | if (!ext4_group_desc_csum_verify(sbi, i, gdp)) { |
b31e1552 ES |
2052 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
2053 | "Checksum for group %u failed (%u!=%u)", | |
2054 | i, le16_to_cpu(ext4_group_desc_csum(sbi, i, | |
2055 | gdp)), le16_to_cpu(gdp->bg_checksum)); | |
7ee1ec4c | 2056 | if (!(sb->s_flags & MS_RDONLY)) { |
955ce5f5 | 2057 | ext4_unlock_group(sb, i); |
8a266467 | 2058 | return 0; |
7ee1ec4c | 2059 | } |
717d50e4 | 2060 | } |
955ce5f5 | 2061 | ext4_unlock_group(sb, i); |
ce421581 JS |
2062 | if (!flexbg_flag) |
2063 | first_block += EXT4_BLOCKS_PER_GROUP(sb); | |
ac27a0ec | 2064 | } |
bfff6873 LC |
2065 | if (NULL != first_not_zeroed) |
2066 | *first_not_zeroed = grp; | |
ac27a0ec | 2067 | |
bd81d8ee | 2068 | ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb)); |
0b8e58a1 | 2069 | sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb)); |
ac27a0ec DK |
2070 | return 1; |
2071 | } | |
2072 | ||
617ba13b | 2073 | /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at |
ac27a0ec DK |
2074 | * the superblock) which were deleted from all directories, but held open by |
2075 | * a process at the time of a crash. We walk the list and try to delete these | |
2076 | * inodes at recovery time (only with a read-write filesystem). | |
2077 | * | |
2078 | * In order to keep the orphan inode chain consistent during traversal (in | |
2079 | * case of crash during recovery), we link each inode into the superblock | |
2080 | * orphan list_head and handle it the same way as an inode deletion during | |
2081 | * normal operation (which journals the operations for us). | |
2082 | * | |
2083 | * We only do an iget() and an iput() on each inode, which is very safe if we | |
2084 | * accidentally point at an in-use or already deleted inode. The worst that | |
2085 | * can happen in this case is that we get a "bit already cleared" message from | |
617ba13b | 2086 | * ext4_free_inode(). The only reason we would point at a wrong inode is if |
ac27a0ec DK |
2087 | * e2fsck was run on this filesystem, and it must have already done the orphan |
2088 | * inode cleanup for us, so we can safely abort without any further action. | |
2089 | */ | |
2b2d6d01 TT |
2090 | static void ext4_orphan_cleanup(struct super_block *sb, |
2091 | struct ext4_super_block *es) | |
ac27a0ec DK |
2092 | { |
2093 | unsigned int s_flags = sb->s_flags; | |
2094 | int nr_orphans = 0, nr_truncates = 0; | |
2095 | #ifdef CONFIG_QUOTA | |
2096 | int i; | |
2097 | #endif | |
2098 | if (!es->s_last_orphan) { | |
2099 | jbd_debug(4, "no orphan inodes to clean up\n"); | |
2100 | return; | |
2101 | } | |
2102 | ||
a8f48a95 | 2103 | if (bdev_read_only(sb->s_bdev)) { |
b31e1552 ES |
2104 | ext4_msg(sb, KERN_ERR, "write access " |
2105 | "unavailable, skipping orphan cleanup"); | |
a8f48a95 ES |
2106 | return; |
2107 | } | |
2108 | ||
617ba13b | 2109 | if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { |
ac27a0ec DK |
2110 | if (es->s_last_orphan) |
2111 | jbd_debug(1, "Errors on filesystem, " | |
2112 | "clearing orphan list.\n"); | |
2113 | es->s_last_orphan = 0; | |
2114 | jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); | |
2115 | return; | |
2116 | } | |
2117 | ||
2118 | if (s_flags & MS_RDONLY) { | |
b31e1552 | 2119 | ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); |
ac27a0ec DK |
2120 | sb->s_flags &= ~MS_RDONLY; |
2121 | } | |
2122 | #ifdef CONFIG_QUOTA | |
2123 | /* Needed for iput() to work correctly and not trash data */ | |
2124 | sb->s_flags |= MS_ACTIVE; | |
2125 | /* Turn on quotas so that they are updated correctly */ | |
2126 | for (i = 0; i < MAXQUOTAS; i++) { | |
617ba13b MC |
2127 | if (EXT4_SB(sb)->s_qf_names[i]) { |
2128 | int ret = ext4_quota_on_mount(sb, i); | |
ac27a0ec | 2129 | if (ret < 0) |
b31e1552 ES |
2130 | ext4_msg(sb, KERN_ERR, |
2131 | "Cannot turn on journaled " | |
2132 | "quota: error %d", ret); | |
ac27a0ec DK |
2133 | } |
2134 | } | |
2135 | #endif | |
2136 | ||
2137 | while (es->s_last_orphan) { | |
2138 | struct inode *inode; | |
2139 | ||
97bd42b9 JB |
2140 | inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); |
2141 | if (IS_ERR(inode)) { | |
ac27a0ec DK |
2142 | es->s_last_orphan = 0; |
2143 | break; | |
2144 | } | |
2145 | ||
617ba13b | 2146 | list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); |
871a2931 | 2147 | dquot_initialize(inode); |
ac27a0ec | 2148 | if (inode->i_nlink) { |
b31e1552 ES |
2149 | ext4_msg(sb, KERN_DEBUG, |
2150 | "%s: truncating inode %lu to %lld bytes", | |
46e665e9 | 2151 | __func__, inode->i_ino, inode->i_size); |
e5f8eab8 | 2152 | jbd_debug(2, "truncating inode %lu to %lld bytes\n", |
ac27a0ec | 2153 | inode->i_ino, inode->i_size); |
617ba13b | 2154 | ext4_truncate(inode); |
ac27a0ec DK |
2155 | nr_truncates++; |
2156 | } else { | |
b31e1552 ES |
2157 | ext4_msg(sb, KERN_DEBUG, |
2158 | "%s: deleting unreferenced inode %lu", | |
46e665e9 | 2159 | __func__, inode->i_ino); |
ac27a0ec DK |
2160 | jbd_debug(2, "deleting unreferenced inode %lu\n", |
2161 | inode->i_ino); | |
2162 | nr_orphans++; | |
2163 | } | |
2164 | iput(inode); /* The delete magic happens here! */ | |
2165 | } | |
2166 | ||
2b2d6d01 | 2167 | #define PLURAL(x) (x), ((x) == 1) ? "" : "s" |
ac27a0ec DK |
2168 | |
2169 | if (nr_orphans) | |
b31e1552 ES |
2170 | ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", |
2171 | PLURAL(nr_orphans)); | |
ac27a0ec | 2172 | if (nr_truncates) |
b31e1552 ES |
2173 | ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", |
2174 | PLURAL(nr_truncates)); | |
ac27a0ec DK |
2175 | #ifdef CONFIG_QUOTA |
2176 | /* Turn quotas off */ | |
2177 | for (i = 0; i < MAXQUOTAS; i++) { | |
2178 | if (sb_dqopt(sb)->files[i]) | |
287a8095 | 2179 | dquot_quota_off(sb, i); |
ac27a0ec DK |
2180 | } |
2181 | #endif | |
2182 | sb->s_flags = s_flags; /* Restore MS_RDONLY status */ | |
2183 | } | |
0b8e58a1 | 2184 | |
cd2291a4 ES |
2185 | /* |
2186 | * Maximal extent format file size. | |
2187 | * Resulting logical blkno at s_maxbytes must fit in our on-disk | |
2188 | * extent format containers, within a sector_t, and within i_blocks | |
2189 | * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, | |
2190 | * so that won't be a limiting factor. | |
2191 | * | |
2192 | * Note, this does *not* consider any metadata overhead for vfs i_blocks. | |
2193 | */ | |
f287a1a5 | 2194 | static loff_t ext4_max_size(int blkbits, int has_huge_files) |
cd2291a4 ES |
2195 | { |
2196 | loff_t res; | |
2197 | loff_t upper_limit = MAX_LFS_FILESIZE; | |
2198 | ||
2199 | /* small i_blocks in vfs inode? */ | |
f287a1a5 | 2200 | if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { |
cd2291a4 | 2201 | /* |
90c699a9 | 2202 | * CONFIG_LBDAF is not enabled implies the inode |
cd2291a4 ES |
2203 | * i_block represent total blocks in 512 bytes |
2204 | * 32 == size of vfs inode i_blocks * 8 | |
2205 | */ | |
2206 | upper_limit = (1LL << 32) - 1; | |
2207 | ||
2208 | /* total blocks in file system block size */ | |
2209 | upper_limit >>= (blkbits - 9); | |
2210 | upper_limit <<= blkbits; | |
2211 | } | |
2212 | ||
2213 | /* 32-bit extent-start container, ee_block */ | |
2214 | res = 1LL << 32; | |
2215 | res <<= blkbits; | |
2216 | res -= 1; | |
2217 | ||
2218 | /* Sanity check against vm- & vfs- imposed limits */ | |
2219 | if (res > upper_limit) | |
2220 | res = upper_limit; | |
2221 | ||
2222 | return res; | |
2223 | } | |
ac27a0ec | 2224 | |
ac27a0ec | 2225 | /* |
cd2291a4 | 2226 | * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect |
0fc1b451 AK |
2227 | * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. |
2228 | * We need to be 1 filesystem block less than the 2^48 sector limit. | |
ac27a0ec | 2229 | */ |
f287a1a5 | 2230 | static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) |
ac27a0ec | 2231 | { |
617ba13b | 2232 | loff_t res = EXT4_NDIR_BLOCKS; |
0fc1b451 AK |
2233 | int meta_blocks; |
2234 | loff_t upper_limit; | |
0b8e58a1 AD |
2235 | /* This is calculated to be the largest file size for a dense, block |
2236 | * mapped file such that the file's total number of 512-byte sectors, | |
2237 | * including data and all indirect blocks, does not exceed (2^48 - 1). | |
2238 | * | |
2239 | * __u32 i_blocks_lo and _u16 i_blocks_high represent the total | |
2240 | * number of 512-byte sectors of the file. | |
0fc1b451 AK |
2241 | */ |
2242 | ||
f287a1a5 | 2243 | if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { |
0fc1b451 | 2244 | /* |
90c699a9 | 2245 | * !has_huge_files or CONFIG_LBDAF not enabled implies that |
0b8e58a1 AD |
2246 | * the inode i_block field represents total file blocks in |
2247 | * 2^32 512-byte sectors == size of vfs inode i_blocks * 8 | |
0fc1b451 AK |
2248 | */ |
2249 | upper_limit = (1LL << 32) - 1; | |
2250 | ||
2251 | /* total blocks in file system block size */ | |
2252 | upper_limit >>= (bits - 9); | |
2253 | ||
2254 | } else { | |
8180a562 AK |
2255 | /* |
2256 | * We use 48 bit ext4_inode i_blocks | |
2257 | * With EXT4_HUGE_FILE_FL set the i_blocks | |
2258 | * represent total number of blocks in | |
2259 | * file system block size | |
2260 | */ | |
0fc1b451 AK |
2261 | upper_limit = (1LL << 48) - 1; |
2262 | ||
0fc1b451 AK |
2263 | } |
2264 | ||
2265 | /* indirect blocks */ | |
2266 | meta_blocks = 1; | |
2267 | /* double indirect blocks */ | |
2268 | meta_blocks += 1 + (1LL << (bits-2)); | |
2269 | /* tripple indirect blocks */ | |
2270 | meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); | |
2271 | ||
2272 | upper_limit -= meta_blocks; | |
2273 | upper_limit <<= bits; | |
ac27a0ec DK |
2274 | |
2275 | res += 1LL << (bits-2); | |
2276 | res += 1LL << (2*(bits-2)); | |
2277 | res += 1LL << (3*(bits-2)); | |
2278 | res <<= bits; | |
2279 | if (res > upper_limit) | |
2280 | res = upper_limit; | |
0fc1b451 AK |
2281 | |
2282 | if (res > MAX_LFS_FILESIZE) | |
2283 | res = MAX_LFS_FILESIZE; | |
2284 | ||
ac27a0ec DK |
2285 | return res; |
2286 | } | |
2287 | ||
617ba13b | 2288 | static ext4_fsblk_t descriptor_loc(struct super_block *sb, |
0b8e58a1 | 2289 | ext4_fsblk_t logical_sb_block, int nr) |
ac27a0ec | 2290 | { |
617ba13b | 2291 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
fd2d4291 | 2292 | ext4_group_t bg, first_meta_bg; |
ac27a0ec DK |
2293 | int has_super = 0; |
2294 | ||
2295 | first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); | |
2296 | ||
617ba13b | 2297 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) || |
ac27a0ec | 2298 | nr < first_meta_bg) |
70bbb3e0 | 2299 | return logical_sb_block + nr + 1; |
ac27a0ec | 2300 | bg = sbi->s_desc_per_block * nr; |
617ba13b | 2301 | if (ext4_bg_has_super(sb, bg)) |
ac27a0ec | 2302 | has_super = 1; |
0b8e58a1 | 2303 | |
617ba13b | 2304 | return (has_super + ext4_group_first_block_no(sb, bg)); |
ac27a0ec DK |
2305 | } |
2306 | ||
c9de560d AT |
2307 | /** |
2308 | * ext4_get_stripe_size: Get the stripe size. | |
2309 | * @sbi: In memory super block info | |
2310 | * | |
2311 | * If we have specified it via mount option, then | |
2312 | * use the mount option value. If the value specified at mount time is | |
2313 | * greater than the blocks per group use the super block value. | |
2314 | * If the super block value is greater than blocks per group return 0. | |
2315 | * Allocator needs it be less than blocks per group. | |
2316 | * | |
2317 | */ | |
2318 | static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) | |
2319 | { | |
2320 | unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); | |
2321 | unsigned long stripe_width = | |
2322 | le32_to_cpu(sbi->s_es->s_raid_stripe_width); | |
2323 | ||
2324 | if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) | |
2325 | return sbi->s_stripe; | |
2326 | ||
2327 | if (stripe_width <= sbi->s_blocks_per_group) | |
2328 | return stripe_width; | |
2329 | ||
2330 | if (stride <= sbi->s_blocks_per_group) | |
2331 | return stride; | |
2332 | ||
2333 | return 0; | |
2334 | } | |
ac27a0ec | 2335 | |
3197ebdb TT |
2336 | /* sysfs supprt */ |
2337 | ||
2338 | struct ext4_attr { | |
2339 | struct attribute attr; | |
2340 | ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *); | |
60e6679e | 2341 | ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *, |
3197ebdb TT |
2342 | const char *, size_t); |
2343 | int offset; | |
2344 | }; | |
2345 | ||
2346 | static int parse_strtoul(const char *buf, | |
2347 | unsigned long max, unsigned long *value) | |
2348 | { | |
2349 | char *endp; | |
2350 | ||
e7d2860b AGR |
2351 | *value = simple_strtoul(skip_spaces(buf), &endp, 0); |
2352 | endp = skip_spaces(endp); | |
3197ebdb TT |
2353 | if (*endp || *value > max) |
2354 | return -EINVAL; | |
2355 | ||
2356 | return 0; | |
2357 | } | |
2358 | ||
2359 | static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a, | |
2360 | struct ext4_sb_info *sbi, | |
2361 | char *buf) | |
2362 | { | |
2363 | return snprintf(buf, PAGE_SIZE, "%llu\n", | |
2364 | (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter)); | |
2365 | } | |
2366 | ||
2367 | static ssize_t session_write_kbytes_show(struct ext4_attr *a, | |
2368 | struct ext4_sb_info *sbi, char *buf) | |
2369 | { | |
2370 | struct super_block *sb = sbi->s_buddy_cache->i_sb; | |
2371 | ||
f613dfcb TT |
2372 | if (!sb->s_bdev->bd_part) |
2373 | return snprintf(buf, PAGE_SIZE, "0\n"); | |
3197ebdb TT |
2374 | return snprintf(buf, PAGE_SIZE, "%lu\n", |
2375 | (part_stat_read(sb->s_bdev->bd_part, sectors[1]) - | |
2376 | sbi->s_sectors_written_start) >> 1); | |
2377 | } | |
2378 | ||
2379 | static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a, | |
2380 | struct ext4_sb_info *sbi, char *buf) | |
2381 | { | |
2382 | struct super_block *sb = sbi->s_buddy_cache->i_sb; | |
2383 | ||
f613dfcb TT |
2384 | if (!sb->s_bdev->bd_part) |
2385 | return snprintf(buf, PAGE_SIZE, "0\n"); | |
3197ebdb | 2386 | return snprintf(buf, PAGE_SIZE, "%llu\n", |
a6b43e38 | 2387 | (unsigned long long)(sbi->s_kbytes_written + |
3197ebdb | 2388 | ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - |
a6b43e38 | 2389 | EXT4_SB(sb)->s_sectors_written_start) >> 1))); |
3197ebdb TT |
2390 | } |
2391 | ||
2392 | static ssize_t inode_readahead_blks_store(struct ext4_attr *a, | |
2393 | struct ext4_sb_info *sbi, | |
2394 | const char *buf, size_t count) | |
2395 | { | |
2396 | unsigned long t; | |
2397 | ||
2398 | if (parse_strtoul(buf, 0x40000000, &t)) | |
2399 | return -EINVAL; | |
2400 | ||
f7c43950 | 2401 | if (!is_power_of_2(t)) |
3197ebdb TT |
2402 | return -EINVAL; |
2403 | ||
2404 | sbi->s_inode_readahead_blks = t; | |
2405 | return count; | |
2406 | } | |
2407 | ||
2408 | static ssize_t sbi_ui_show(struct ext4_attr *a, | |
0b8e58a1 | 2409 | struct ext4_sb_info *sbi, char *buf) |
3197ebdb TT |
2410 | { |
2411 | unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset); | |
2412 | ||
2413 | return snprintf(buf, PAGE_SIZE, "%u\n", *ui); | |
2414 | } | |
2415 | ||
2416 | static ssize_t sbi_ui_store(struct ext4_attr *a, | |
2417 | struct ext4_sb_info *sbi, | |
2418 | const char *buf, size_t count) | |
2419 | { | |
2420 | unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset); | |
2421 | unsigned long t; | |
2422 | ||
2423 | if (parse_strtoul(buf, 0xffffffff, &t)) | |
2424 | return -EINVAL; | |
2425 | *ui = t; | |
2426 | return count; | |
2427 | } | |
2428 | ||
2429 | #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \ | |
2430 | static struct ext4_attr ext4_attr_##_name = { \ | |
2431 | .attr = {.name = __stringify(_name), .mode = _mode }, \ | |
2432 | .show = _show, \ | |
2433 | .store = _store, \ | |
2434 | .offset = offsetof(struct ext4_sb_info, _elname), \ | |
2435 | } | |
2436 | #define EXT4_ATTR(name, mode, show, store) \ | |
2437 | static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store) | |
2438 | ||
857ac889 | 2439 | #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL) |
3197ebdb TT |
2440 | #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL) |
2441 | #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store) | |
2442 | #define EXT4_RW_ATTR_SBI_UI(name, elname) \ | |
2443 | EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname) | |
2444 | #define ATTR_LIST(name) &ext4_attr_##name.attr | |
2445 | ||
2446 | EXT4_RO_ATTR(delayed_allocation_blocks); | |
2447 | EXT4_RO_ATTR(session_write_kbytes); | |
2448 | EXT4_RO_ATTR(lifetime_write_kbytes); | |
2449 | EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show, | |
2450 | inode_readahead_blks_store, s_inode_readahead_blks); | |
11013911 | 2451 | EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal); |
3197ebdb TT |
2452 | EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats); |
2453 | EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan); | |
2454 | EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan); | |
2455 | EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs); | |
2456 | EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request); | |
2457 | EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc); | |
55138e0b | 2458 | EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump); |
3197ebdb TT |
2459 | |
2460 | static struct attribute *ext4_attrs[] = { | |
2461 | ATTR_LIST(delayed_allocation_blocks), | |
2462 | ATTR_LIST(session_write_kbytes), | |
2463 | ATTR_LIST(lifetime_write_kbytes), | |
2464 | ATTR_LIST(inode_readahead_blks), | |
11013911 | 2465 | ATTR_LIST(inode_goal), |
3197ebdb TT |
2466 | ATTR_LIST(mb_stats), |
2467 | ATTR_LIST(mb_max_to_scan), | |
2468 | ATTR_LIST(mb_min_to_scan), | |
2469 | ATTR_LIST(mb_order2_req), | |
2470 | ATTR_LIST(mb_stream_req), | |
2471 | ATTR_LIST(mb_group_prealloc), | |
55138e0b | 2472 | ATTR_LIST(max_writeback_mb_bump), |
3197ebdb TT |
2473 | NULL, |
2474 | }; | |
2475 | ||
857ac889 LC |
2476 | /* Features this copy of ext4 supports */ |
2477 | EXT4_INFO_ATTR(lazy_itable_init); | |
27ee40df | 2478 | EXT4_INFO_ATTR(batched_discard); |
857ac889 LC |
2479 | |
2480 | static struct attribute *ext4_feat_attrs[] = { | |
2481 | ATTR_LIST(lazy_itable_init), | |
27ee40df | 2482 | ATTR_LIST(batched_discard), |
857ac889 LC |
2483 | NULL, |
2484 | }; | |
2485 | ||
3197ebdb TT |
2486 | static ssize_t ext4_attr_show(struct kobject *kobj, |
2487 | struct attribute *attr, char *buf) | |
2488 | { | |
2489 | struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, | |
2490 | s_kobj); | |
2491 | struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); | |
2492 | ||
2493 | return a->show ? a->show(a, sbi, buf) : 0; | |
2494 | } | |
2495 | ||
2496 | static ssize_t ext4_attr_store(struct kobject *kobj, | |
2497 | struct attribute *attr, | |
2498 | const char *buf, size_t len) | |
2499 | { | |
2500 | struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, | |
2501 | s_kobj); | |
2502 | struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); | |
2503 | ||
2504 | return a->store ? a->store(a, sbi, buf, len) : 0; | |
2505 | } | |
2506 | ||
2507 | static void ext4_sb_release(struct kobject *kobj) | |
2508 | { | |
2509 | struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, | |
2510 | s_kobj); | |
2511 | complete(&sbi->s_kobj_unregister); | |
2512 | } | |
2513 | ||
52cf25d0 | 2514 | static const struct sysfs_ops ext4_attr_ops = { |
3197ebdb TT |
2515 | .show = ext4_attr_show, |
2516 | .store = ext4_attr_store, | |
2517 | }; | |
2518 | ||
2519 | static struct kobj_type ext4_ktype = { | |
2520 | .default_attrs = ext4_attrs, | |
2521 | .sysfs_ops = &ext4_attr_ops, | |
2522 | .release = ext4_sb_release, | |
2523 | }; | |
2524 | ||
857ac889 LC |
2525 | static void ext4_feat_release(struct kobject *kobj) |
2526 | { | |
2527 | complete(&ext4_feat->f_kobj_unregister); | |
2528 | } | |
2529 | ||
2530 | static struct kobj_type ext4_feat_ktype = { | |
2531 | .default_attrs = ext4_feat_attrs, | |
2532 | .sysfs_ops = &ext4_attr_ops, | |
2533 | .release = ext4_feat_release, | |
2534 | }; | |
2535 | ||
a13fb1a4 ES |
2536 | /* |
2537 | * Check whether this filesystem can be mounted based on | |
2538 | * the features present and the RDONLY/RDWR mount requested. | |
2539 | * Returns 1 if this filesystem can be mounted as requested, | |
2540 | * 0 if it cannot be. | |
2541 | */ | |
2542 | static int ext4_feature_set_ok(struct super_block *sb, int readonly) | |
2543 | { | |
2544 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) { | |
2545 | ext4_msg(sb, KERN_ERR, | |
2546 | "Couldn't mount because of " | |
2547 | "unsupported optional features (%x)", | |
2548 | (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & | |
2549 | ~EXT4_FEATURE_INCOMPAT_SUPP)); | |
2550 | return 0; | |
2551 | } | |
2552 | ||
2553 | if (readonly) | |
2554 | return 1; | |
2555 | ||
2556 | /* Check that feature set is OK for a read-write mount */ | |
2557 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) { | |
2558 | ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " | |
2559 | "unsupported optional features (%x)", | |
2560 | (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & | |
2561 | ~EXT4_FEATURE_RO_COMPAT_SUPP)); | |
2562 | return 0; | |
2563 | } | |
2564 | /* | |
2565 | * Large file size enabled file system can only be mounted | |
2566 | * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF | |
2567 | */ | |
2568 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) { | |
2569 | if (sizeof(blkcnt_t) < sizeof(u64)) { | |
2570 | ext4_msg(sb, KERN_ERR, "Filesystem with huge files " | |
2571 | "cannot be mounted RDWR without " | |
2572 | "CONFIG_LBDAF"); | |
2573 | return 0; | |
2574 | } | |
2575 | } | |
2576 | return 1; | |
2577 | } | |
2578 | ||
66e61a9e TT |
2579 | /* |
2580 | * This function is called once a day if we have errors logged | |
2581 | * on the file system | |
2582 | */ | |
2583 | static void print_daily_error_info(unsigned long arg) | |
2584 | { | |
2585 | struct super_block *sb = (struct super_block *) arg; | |
2586 | struct ext4_sb_info *sbi; | |
2587 | struct ext4_super_block *es; | |
2588 | ||
2589 | sbi = EXT4_SB(sb); | |
2590 | es = sbi->s_es; | |
2591 | ||
2592 | if (es->s_error_count) | |
2593 | ext4_msg(sb, KERN_NOTICE, "error count: %u", | |
2594 | le32_to_cpu(es->s_error_count)); | |
2595 | if (es->s_first_error_time) { | |
2596 | printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d", | |
2597 | sb->s_id, le32_to_cpu(es->s_first_error_time), | |
2598 | (int) sizeof(es->s_first_error_func), | |
2599 | es->s_first_error_func, | |
2600 | le32_to_cpu(es->s_first_error_line)); | |
2601 | if (es->s_first_error_ino) | |
2602 | printk(": inode %u", | |
2603 | le32_to_cpu(es->s_first_error_ino)); | |
2604 | if (es->s_first_error_block) | |
2605 | printk(": block %llu", (unsigned long long) | |
2606 | le64_to_cpu(es->s_first_error_block)); | |
2607 | printk("\n"); | |
2608 | } | |
2609 | if (es->s_last_error_time) { | |
2610 | printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d", | |
2611 | sb->s_id, le32_to_cpu(es->s_last_error_time), | |
2612 | (int) sizeof(es->s_last_error_func), | |
2613 | es->s_last_error_func, | |
2614 | le32_to_cpu(es->s_last_error_line)); | |
2615 | if (es->s_last_error_ino) | |
2616 | printk(": inode %u", | |
2617 | le32_to_cpu(es->s_last_error_ino)); | |
2618 | if (es->s_last_error_block) | |
2619 | printk(": block %llu", (unsigned long long) | |
2620 | le64_to_cpu(es->s_last_error_block)); | |
2621 | printk("\n"); | |
2622 | } | |
2623 | mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ | |
2624 | } | |
2625 | ||
bfff6873 LC |
2626 | static void ext4_lazyinode_timeout(unsigned long data) |
2627 | { | |
2628 | struct task_struct *p = (struct task_struct *)data; | |
2629 | wake_up_process(p); | |
2630 | } | |
2631 | ||
2632 | /* Find next suitable group and run ext4_init_inode_table */ | |
2633 | static int ext4_run_li_request(struct ext4_li_request *elr) | |
2634 | { | |
2635 | struct ext4_group_desc *gdp = NULL; | |
2636 | ext4_group_t group, ngroups; | |
2637 | struct super_block *sb; | |
2638 | unsigned long timeout = 0; | |
2639 | int ret = 0; | |
2640 | ||
2641 | sb = elr->lr_super; | |
2642 | ngroups = EXT4_SB(sb)->s_groups_count; | |
2643 | ||
2644 | for (group = elr->lr_next_group; group < ngroups; group++) { | |
2645 | gdp = ext4_get_group_desc(sb, group, NULL); | |
2646 | if (!gdp) { | |
2647 | ret = 1; | |
2648 | break; | |
2649 | } | |
2650 | ||
2651 | if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) | |
2652 | break; | |
2653 | } | |
2654 | ||
2655 | if (group == ngroups) | |
2656 | ret = 1; | |
2657 | ||
2658 | if (!ret) { | |
2659 | timeout = jiffies; | |
2660 | ret = ext4_init_inode_table(sb, group, | |
2661 | elr->lr_timeout ? 0 : 1); | |
2662 | if (elr->lr_timeout == 0) { | |
2663 | timeout = jiffies - timeout; | |
2664 | if (elr->lr_sbi->s_li_wait_mult) | |
2665 | timeout *= elr->lr_sbi->s_li_wait_mult; | |
2666 | else | |
2667 | timeout *= 20; | |
2668 | elr->lr_timeout = timeout; | |
2669 | } | |
2670 | elr->lr_next_sched = jiffies + elr->lr_timeout; | |
2671 | elr->lr_next_group = group + 1; | |
2672 | } | |
2673 | ||
2674 | return ret; | |
2675 | } | |
2676 | ||
2677 | /* | |
2678 | * Remove lr_request from the list_request and free the | |
2679 | * request tructure. Should be called with li_list_mtx held | |
2680 | */ | |
2681 | static void ext4_remove_li_request(struct ext4_li_request *elr) | |
2682 | { | |
2683 | struct ext4_sb_info *sbi; | |
2684 | ||
2685 | if (!elr) | |
2686 | return; | |
2687 | ||
2688 | sbi = elr->lr_sbi; | |
2689 | ||
2690 | list_del(&elr->lr_request); | |
2691 | sbi->s_li_request = NULL; | |
2692 | kfree(elr); | |
2693 | } | |
2694 | ||
2695 | static void ext4_unregister_li_request(struct super_block *sb) | |
2696 | { | |
2697 | struct ext4_li_request *elr = EXT4_SB(sb)->s_li_request; | |
2698 | ||
2699 | if (!ext4_li_info) | |
2700 | return; | |
2701 | ||
2702 | mutex_lock(&ext4_li_info->li_list_mtx); | |
2703 | ext4_remove_li_request(elr); | |
2704 | mutex_unlock(&ext4_li_info->li_list_mtx); | |
2705 | } | |
2706 | ||
2707 | /* | |
2708 | * This is the function where ext4lazyinit thread lives. It walks | |
2709 | * through the request list searching for next scheduled filesystem. | |
2710 | * When such a fs is found, run the lazy initialization request | |
2711 | * (ext4_rn_li_request) and keep track of the time spend in this | |
2712 | * function. Based on that time we compute next schedule time of | |
2713 | * the request. When walking through the list is complete, compute | |
2714 | * next waking time and put itself into sleep. | |
2715 | */ | |
2716 | static int ext4_lazyinit_thread(void *arg) | |
2717 | { | |
2718 | struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; | |
2719 | struct list_head *pos, *n; | |
2720 | struct ext4_li_request *elr; | |
2721 | unsigned long next_wakeup; | |
2722 | DEFINE_WAIT(wait); | |
bfff6873 LC |
2723 | |
2724 | BUG_ON(NULL == eli); | |
2725 | ||
2726 | eli->li_timer.data = (unsigned long)current; | |
2727 | eli->li_timer.function = ext4_lazyinode_timeout; | |
2728 | ||
2729 | eli->li_task = current; | |
2730 | wake_up(&eli->li_wait_task); | |
2731 | ||
2732 | cont_thread: | |
2733 | while (true) { | |
2734 | next_wakeup = MAX_JIFFY_OFFSET; | |
2735 | ||
2736 | mutex_lock(&eli->li_list_mtx); | |
2737 | if (list_empty(&eli->li_request_list)) { | |
2738 | mutex_unlock(&eli->li_list_mtx); | |
2739 | goto exit_thread; | |
2740 | } | |
2741 | ||
2742 | list_for_each_safe(pos, n, &eli->li_request_list) { | |
2743 | elr = list_entry(pos, struct ext4_li_request, | |
2744 | lr_request); | |
2745 | ||
b2c78cd0 TT |
2746 | if (time_after_eq(jiffies, elr->lr_next_sched)) { |
2747 | if (ext4_run_li_request(elr) != 0) { | |
2748 | /* error, remove the lazy_init job */ | |
2749 | ext4_remove_li_request(elr); | |
2750 | continue; | |
2751 | } | |
bfff6873 LC |
2752 | } |
2753 | ||
2754 | if (time_before(elr->lr_next_sched, next_wakeup)) | |
2755 | next_wakeup = elr->lr_next_sched; | |
2756 | } | |
2757 | mutex_unlock(&eli->li_list_mtx); | |
2758 | ||
2759 | if (freezing(current)) | |
2760 | refrigerator(); | |
2761 | ||
f4245bd4 LC |
2762 | if ((time_after_eq(jiffies, next_wakeup)) || |
2763 | (MAX_JIFFY_OFFSET == next_wakeup)) { | |
bfff6873 LC |
2764 | cond_resched(); |
2765 | continue; | |
2766 | } | |
2767 | ||
2768 | eli->li_timer.expires = next_wakeup; | |
2769 | add_timer(&eli->li_timer); | |
2770 | prepare_to_wait(&eli->li_wait_daemon, &wait, | |
2771 | TASK_INTERRUPTIBLE); | |
2772 | if (time_before(jiffies, next_wakeup)) | |
2773 | schedule(); | |
2774 | finish_wait(&eli->li_wait_daemon, &wait); | |
2775 | } | |
2776 | ||
2777 | exit_thread: | |
2778 | /* | |
2779 | * It looks like the request list is empty, but we need | |
2780 | * to check it under the li_list_mtx lock, to prevent any | |
2781 | * additions into it, and of course we should lock ext4_li_mtx | |
2782 | * to atomically free the list and ext4_li_info, because at | |
2783 | * this point another ext4 filesystem could be registering | |
2784 | * new one. | |
2785 | */ | |
2786 | mutex_lock(&ext4_li_mtx); | |
2787 | mutex_lock(&eli->li_list_mtx); | |
2788 | if (!list_empty(&eli->li_request_list)) { | |
2789 | mutex_unlock(&eli->li_list_mtx); | |
2790 | mutex_unlock(&ext4_li_mtx); | |
2791 | goto cont_thread; | |
2792 | } | |
2793 | mutex_unlock(&eli->li_list_mtx); | |
2794 | del_timer_sync(&ext4_li_info->li_timer); | |
2795 | eli->li_task = NULL; | |
2796 | wake_up(&eli->li_wait_task); | |
2797 | ||
2798 | kfree(ext4_li_info); | |
2799 | ext4_li_info = NULL; | |
2800 | mutex_unlock(&ext4_li_mtx); | |
2801 | ||
2802 | return 0; | |
2803 | } | |
2804 | ||
2805 | static void ext4_clear_request_list(void) | |
2806 | { | |
2807 | struct list_head *pos, *n; | |
2808 | struct ext4_li_request *elr; | |
2809 | ||
2810 | mutex_lock(&ext4_li_info->li_list_mtx); | |
bfff6873 LC |
2811 | list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { |
2812 | elr = list_entry(pos, struct ext4_li_request, | |
2813 | lr_request); | |
2814 | ext4_remove_li_request(elr); | |
2815 | } | |
2816 | mutex_unlock(&ext4_li_info->li_list_mtx); | |
2817 | } | |
2818 | ||
2819 | static int ext4_run_lazyinit_thread(void) | |
2820 | { | |
2821 | struct task_struct *t; | |
2822 | ||
2823 | t = kthread_run(ext4_lazyinit_thread, ext4_li_info, "ext4lazyinit"); | |
2824 | if (IS_ERR(t)) { | |
2825 | int err = PTR_ERR(t); | |
2826 | ext4_clear_request_list(); | |
2827 | del_timer_sync(&ext4_li_info->li_timer); | |
2828 | kfree(ext4_li_info); | |
2829 | ext4_li_info = NULL; | |
2830 | printk(KERN_CRIT "EXT4: error %d creating inode table " | |
2831 | "initialization thread\n", | |
2832 | err); | |
2833 | return err; | |
2834 | } | |
2835 | ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; | |
2836 | ||
2837 | wait_event(ext4_li_info->li_wait_task, ext4_li_info->li_task != NULL); | |
2838 | return 0; | |
2839 | } | |
2840 | ||
2841 | /* | |
2842 | * Check whether it make sense to run itable init. thread or not. | |
2843 | * If there is at least one uninitialized inode table, return | |
2844 | * corresponding group number, else the loop goes through all | |
2845 | * groups and return total number of groups. | |
2846 | */ | |
2847 | static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) | |
2848 | { | |
2849 | ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; | |
2850 | struct ext4_group_desc *gdp = NULL; | |
2851 | ||
2852 | for (group = 0; group < ngroups; group++) { | |
2853 | gdp = ext4_get_group_desc(sb, group, NULL); | |
2854 | if (!gdp) | |
2855 | continue; | |
2856 | ||
2857 | if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) | |
2858 | break; | |
2859 | } | |
2860 | ||
2861 | return group; | |
2862 | } | |
2863 | ||
2864 | static int ext4_li_info_new(void) | |
2865 | { | |
2866 | struct ext4_lazy_init *eli = NULL; | |
2867 | ||
2868 | eli = kzalloc(sizeof(*eli), GFP_KERNEL); | |
2869 | if (!eli) | |
2870 | return -ENOMEM; | |
2871 | ||
2872 | eli->li_task = NULL; | |
2873 | INIT_LIST_HEAD(&eli->li_request_list); | |
2874 | mutex_init(&eli->li_list_mtx); | |
2875 | ||
2876 | init_waitqueue_head(&eli->li_wait_daemon); | |
2877 | init_waitqueue_head(&eli->li_wait_task); | |
2878 | init_timer(&eli->li_timer); | |
2879 | eli->li_state |= EXT4_LAZYINIT_QUIT; | |
2880 | ||
2881 | ext4_li_info = eli; | |
2882 | ||
2883 | return 0; | |
2884 | } | |
2885 | ||
2886 | static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, | |
2887 | ext4_group_t start) | |
2888 | { | |
2889 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
2890 | struct ext4_li_request *elr; | |
2891 | unsigned long rnd; | |
2892 | ||
2893 | elr = kzalloc(sizeof(*elr), GFP_KERNEL); | |
2894 | if (!elr) | |
2895 | return NULL; | |
2896 | ||
2897 | elr->lr_super = sb; | |
2898 | elr->lr_sbi = sbi; | |
2899 | elr->lr_next_group = start; | |
2900 | ||
2901 | /* | |
2902 | * Randomize first schedule time of the request to | |
2903 | * spread the inode table initialization requests | |
2904 | * better. | |
2905 | */ | |
2906 | get_random_bytes(&rnd, sizeof(rnd)); | |
2907 | elr->lr_next_sched = jiffies + (unsigned long)rnd % | |
2908 | (EXT4_DEF_LI_MAX_START_DELAY * HZ); | |
2909 | ||
2910 | return elr; | |
2911 | } | |
2912 | ||
2913 | static int ext4_register_li_request(struct super_block *sb, | |
2914 | ext4_group_t first_not_zeroed) | |
2915 | { | |
2916 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
2917 | struct ext4_li_request *elr; | |
2918 | ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; | |
beed5ecb | 2919 | int ret; |
bfff6873 LC |
2920 | |
2921 | if (sbi->s_li_request != NULL) | |
beed5ecb | 2922 | return 0; |
bfff6873 LC |
2923 | |
2924 | if (first_not_zeroed == ngroups || | |
2925 | (sb->s_flags & MS_RDONLY) || | |
2926 | !test_opt(sb, INIT_INODE_TABLE)) { | |
2927 | sbi->s_li_request = NULL; | |
beed5ecb | 2928 | return 0; |
bfff6873 LC |
2929 | } |
2930 | ||
2931 | if (first_not_zeroed == ngroups) { | |
2932 | sbi->s_li_request = NULL; | |
beed5ecb | 2933 | return 0; |
bfff6873 LC |
2934 | } |
2935 | ||
2936 | elr = ext4_li_request_new(sb, first_not_zeroed); | |
beed5ecb NK |
2937 | if (!elr) |
2938 | return -ENOMEM; | |
bfff6873 LC |
2939 | |
2940 | mutex_lock(&ext4_li_mtx); | |
2941 | ||
2942 | if (NULL == ext4_li_info) { | |
2943 | ret = ext4_li_info_new(); | |
2944 | if (ret) | |
2945 | goto out; | |
2946 | } | |
2947 | ||
2948 | mutex_lock(&ext4_li_info->li_list_mtx); | |
2949 | list_add(&elr->lr_request, &ext4_li_info->li_request_list); | |
2950 | mutex_unlock(&ext4_li_info->li_list_mtx); | |
2951 | ||
2952 | sbi->s_li_request = elr; | |
2953 | ||
2954 | if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { | |
2955 | ret = ext4_run_lazyinit_thread(); | |
2956 | if (ret) | |
2957 | goto out; | |
2958 | } | |
bfff6873 | 2959 | out: |
beed5ecb NK |
2960 | mutex_unlock(&ext4_li_mtx); |
2961 | if (ret) | |
bfff6873 | 2962 | kfree(elr); |
bfff6873 LC |
2963 | return ret; |
2964 | } | |
2965 | ||
2966 | /* | |
2967 | * We do not need to lock anything since this is called on | |
2968 | * module unload. | |
2969 | */ | |
2970 | static void ext4_destroy_lazyinit_thread(void) | |
2971 | { | |
2972 | /* | |
2973 | * If thread exited earlier | |
2974 | * there's nothing to be done. | |
2975 | */ | |
2976 | if (!ext4_li_info) | |
2977 | return; | |
2978 | ||
2979 | ext4_clear_request_list(); | |
2980 | ||
2981 | while (ext4_li_info->li_task) { | |
2982 | wake_up(&ext4_li_info->li_wait_daemon); | |
2983 | wait_event(ext4_li_info->li_wait_task, | |
2984 | ext4_li_info->li_task == NULL); | |
2985 | } | |
2986 | } | |
2987 | ||
2b2d6d01 | 2988 | static int ext4_fill_super(struct super_block *sb, void *data, int silent) |
7477827f AK |
2989 | __releases(kernel_lock) |
2990 | __acquires(kernel_lock) | |
ac27a0ec | 2991 | { |
d4c402d9 | 2992 | char *orig_data = kstrdup(data, GFP_KERNEL); |
2b2d6d01 | 2993 | struct buffer_head *bh; |
617ba13b MC |
2994 | struct ext4_super_block *es = NULL; |
2995 | struct ext4_sb_info *sbi; | |
2996 | ext4_fsblk_t block; | |
2997 | ext4_fsblk_t sb_block = get_sb_block(&data); | |
70bbb3e0 | 2998 | ext4_fsblk_t logical_sb_block; |
ac27a0ec | 2999 | unsigned long offset = 0; |
ac27a0ec DK |
3000 | unsigned long journal_devnum = 0; |
3001 | unsigned long def_mount_opts; | |
3002 | struct inode *root; | |
9f6200bb | 3003 | char *cp; |
0390131b | 3004 | const char *descr; |
dcc7dae3 | 3005 | int ret = -ENOMEM; |
ac27a0ec | 3006 | int blocksize; |
4ec11028 TT |
3007 | unsigned int db_count; |
3008 | unsigned int i; | |
f287a1a5 | 3009 | int needs_recovery, has_huge_files; |
bd81d8ee | 3010 | __u64 blocks_count; |
833f4077 | 3011 | int err; |
b3881f74 | 3012 | unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; |
bfff6873 | 3013 | ext4_group_t first_not_zeroed; |
ac27a0ec DK |
3014 | |
3015 | sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); | |
3016 | if (!sbi) | |
dcc7dae3 | 3017 | goto out_free_orig; |
705895b6 PE |
3018 | |
3019 | sbi->s_blockgroup_lock = | |
3020 | kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); | |
3021 | if (!sbi->s_blockgroup_lock) { | |
3022 | kfree(sbi); | |
dcc7dae3 | 3023 | goto out_free_orig; |
705895b6 | 3024 | } |
ac27a0ec DK |
3025 | sb->s_fs_info = sbi; |
3026 | sbi->s_mount_opt = 0; | |
617ba13b MC |
3027 | sbi->s_resuid = EXT4_DEF_RESUID; |
3028 | sbi->s_resgid = EXT4_DEF_RESGID; | |
240799cd | 3029 | sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; |
d9c9bef1 | 3030 | sbi->s_sb_block = sb_block; |
f613dfcb TT |
3031 | if (sb->s_bdev->bd_part) |
3032 | sbi->s_sectors_written_start = | |
3033 | part_stat_read(sb->s_bdev->bd_part, sectors[1]); | |
ac27a0ec | 3034 | |
9f6200bb TT |
3035 | /* Cleanup superblock name */ |
3036 | for (cp = sb->s_id; (cp = strchr(cp, '/'));) | |
3037 | *cp = '!'; | |
3038 | ||
dcc7dae3 | 3039 | ret = -EINVAL; |
617ba13b | 3040 | blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); |
ac27a0ec | 3041 | if (!blocksize) { |
b31e1552 | 3042 | ext4_msg(sb, KERN_ERR, "unable to set blocksize"); |
ac27a0ec DK |
3043 | goto out_fail; |
3044 | } | |
3045 | ||
3046 | /* | |
617ba13b | 3047 | * The ext4 superblock will not be buffer aligned for other than 1kB |
ac27a0ec DK |
3048 | * block sizes. We need to calculate the offset from buffer start. |
3049 | */ | |
617ba13b | 3050 | if (blocksize != EXT4_MIN_BLOCK_SIZE) { |
70bbb3e0 AM |
3051 | logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; |
3052 | offset = do_div(logical_sb_block, blocksize); | |
ac27a0ec | 3053 | } else { |
70bbb3e0 | 3054 | logical_sb_block = sb_block; |
ac27a0ec DK |
3055 | } |
3056 | ||
70bbb3e0 | 3057 | if (!(bh = sb_bread(sb, logical_sb_block))) { |
b31e1552 | 3058 | ext4_msg(sb, KERN_ERR, "unable to read superblock"); |
ac27a0ec DK |
3059 | goto out_fail; |
3060 | } | |
3061 | /* | |
3062 | * Note: s_es must be initialized as soon as possible because | |
617ba13b | 3063 | * some ext4 macro-instructions depend on its value |
ac27a0ec | 3064 | */ |
617ba13b | 3065 | es = (struct ext4_super_block *) (((char *)bh->b_data) + offset); |
ac27a0ec DK |
3066 | sbi->s_es = es; |
3067 | sb->s_magic = le16_to_cpu(es->s_magic); | |
617ba13b MC |
3068 | if (sb->s_magic != EXT4_SUPER_MAGIC) |
3069 | goto cantfind_ext4; | |
afc32f7e | 3070 | sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); |
ac27a0ec DK |
3071 | |
3072 | /* Set defaults before we parse the mount options */ | |
3073 | def_mount_opts = le32_to_cpu(es->s_default_mount_opts); | |
bfff6873 | 3074 | set_opt(sbi->s_mount_opt, INIT_INODE_TABLE); |
617ba13b | 3075 | if (def_mount_opts & EXT4_DEFM_DEBUG) |
ac27a0ec | 3076 | set_opt(sbi->s_mount_opt, DEBUG); |
437ca0fd DM |
3077 | if (def_mount_opts & EXT4_DEFM_BSDGROUPS) { |
3078 | ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups", | |
3079 | "2.6.38"); | |
ac27a0ec | 3080 | set_opt(sbi->s_mount_opt, GRPID); |
437ca0fd | 3081 | } |
617ba13b | 3082 | if (def_mount_opts & EXT4_DEFM_UID16) |
ac27a0ec | 3083 | set_opt(sbi->s_mount_opt, NO_UID32); |
03010a33 | 3084 | #ifdef CONFIG_EXT4_FS_XATTR |
617ba13b | 3085 | if (def_mount_opts & EXT4_DEFM_XATTR_USER) |
ac27a0ec | 3086 | set_opt(sbi->s_mount_opt, XATTR_USER); |
2e7842b8 | 3087 | #endif |
03010a33 | 3088 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
617ba13b | 3089 | if (def_mount_opts & EXT4_DEFM_ACL) |
ac27a0ec | 3090 | set_opt(sbi->s_mount_opt, POSIX_ACL); |
2e7842b8 | 3091 | #endif |
617ba13b | 3092 | if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) |
482a7425 | 3093 | set_opt(sbi->s_mount_opt, JOURNAL_DATA); |
617ba13b | 3094 | else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) |
482a7425 | 3095 | set_opt(sbi->s_mount_opt, ORDERED_DATA); |
617ba13b | 3096 | else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) |
482a7425 | 3097 | set_opt(sbi->s_mount_opt, WRITEBACK_DATA); |
617ba13b MC |
3098 | |
3099 | if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) | |
ac27a0ec | 3100 | set_opt(sbi->s_mount_opt, ERRORS_PANIC); |
bb4f397a | 3101 | else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) |
ceea16bf | 3102 | set_opt(sbi->s_mount_opt, ERRORS_CONT); |
bb4f397a AK |
3103 | else |
3104 | set_opt(sbi->s_mount_opt, ERRORS_RO); | |
8b67f04a TT |
3105 | if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY) |
3106 | set_opt(sbi->s_mount_opt, BLOCK_VALIDITY); | |
3107 | if (def_mount_opts & EXT4_DEFM_DISCARD) | |
3108 | set_opt(sbi->s_mount_opt, DISCARD); | |
ac27a0ec DK |
3109 | |
3110 | sbi->s_resuid = le16_to_cpu(es->s_def_resuid); | |
3111 | sbi->s_resgid = le16_to_cpu(es->s_def_resgid); | |
30773840 TT |
3112 | sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; |
3113 | sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; | |
3114 | sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; | |
ac27a0ec | 3115 | |
8b67f04a TT |
3116 | if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) |
3117 | set_opt(sbi->s_mount_opt, BARRIER); | |
ac27a0ec | 3118 | |
dd919b98 AK |
3119 | /* |
3120 | * enable delayed allocation by default | |
3121 | * Use -o nodelalloc to turn it off | |
3122 | */ | |
8b67f04a TT |
3123 | if (!IS_EXT3_SB(sb) && |
3124 | ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) | |
ba69f9ab | 3125 | set_opt(sbi->s_mount_opt, DELALLOC); |
dd919b98 | 3126 | |
8b67f04a TT |
3127 | if (!parse_options((char *) sbi->s_es->s_mount_opts, sb, |
3128 | &journal_devnum, &journal_ioprio, NULL, 0)) { | |
3129 | ext4_msg(sb, KERN_WARNING, | |
3130 | "failed to parse options in superblock: %s", | |
3131 | sbi->s_es->s_mount_opts); | |
3132 | } | |
b3881f74 TT |
3133 | if (!parse_options((char *) data, sb, &journal_devnum, |
3134 | &journal_ioprio, NULL, 0)) | |
ac27a0ec DK |
3135 | goto failed_mount; |
3136 | ||
3137 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
482a7425 | 3138 | (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); |
ac27a0ec | 3139 | |
617ba13b MC |
3140 | if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && |
3141 | (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) || | |
3142 | EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) || | |
3143 | EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U))) | |
b31e1552 ES |
3144 | ext4_msg(sb, KERN_WARNING, |
3145 | "feature flags set on rev 0 fs, " | |
3146 | "running e2fsck is recommended"); | |
469108ff | 3147 | |
ac27a0ec DK |
3148 | /* |
3149 | * Check feature flags regardless of the revision level, since we | |
3150 | * previously didn't change the revision level when setting the flags, | |
3151 | * so there is a chance incompat flags are set on a rev 0 filesystem. | |
3152 | */ | |
a13fb1a4 | 3153 | if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY))) |
ac27a0ec | 3154 | goto failed_mount; |
a13fb1a4 | 3155 | |
ac27a0ec DK |
3156 | blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); |
3157 | ||
617ba13b MC |
3158 | if (blocksize < EXT4_MIN_BLOCK_SIZE || |
3159 | blocksize > EXT4_MAX_BLOCK_SIZE) { | |
b31e1552 ES |
3160 | ext4_msg(sb, KERN_ERR, |
3161 | "Unsupported filesystem blocksize %d", blocksize); | |
ac27a0ec DK |
3162 | goto failed_mount; |
3163 | } | |
3164 | ||
ac27a0ec | 3165 | if (sb->s_blocksize != blocksize) { |
ce40733c AK |
3166 | /* Validate the filesystem blocksize */ |
3167 | if (!sb_set_blocksize(sb, blocksize)) { | |
b31e1552 | 3168 | ext4_msg(sb, KERN_ERR, "bad block size %d", |
ce40733c | 3169 | blocksize); |
ac27a0ec DK |
3170 | goto failed_mount; |
3171 | } | |
3172 | ||
2b2d6d01 | 3173 | brelse(bh); |
70bbb3e0 AM |
3174 | logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; |
3175 | offset = do_div(logical_sb_block, blocksize); | |
3176 | bh = sb_bread(sb, logical_sb_block); | |
ac27a0ec | 3177 | if (!bh) { |
b31e1552 ES |
3178 | ext4_msg(sb, KERN_ERR, |
3179 | "Can't read superblock on 2nd try"); | |
ac27a0ec DK |
3180 | goto failed_mount; |
3181 | } | |
617ba13b | 3182 | es = (struct ext4_super_block *)(((char *)bh->b_data) + offset); |
ac27a0ec | 3183 | sbi->s_es = es; |
617ba13b | 3184 | if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { |
b31e1552 ES |
3185 | ext4_msg(sb, KERN_ERR, |
3186 | "Magic mismatch, very weird!"); | |
ac27a0ec DK |
3187 | goto failed_mount; |
3188 | } | |
3189 | } | |
3190 | ||
a13fb1a4 ES |
3191 | has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb, |
3192 | EXT4_FEATURE_RO_COMPAT_HUGE_FILE); | |
f287a1a5 TT |
3193 | sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, |
3194 | has_huge_files); | |
3195 | sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); | |
ac27a0ec | 3196 | |
617ba13b MC |
3197 | if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { |
3198 | sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; | |
3199 | sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; | |
ac27a0ec DK |
3200 | } else { |
3201 | sbi->s_inode_size = le16_to_cpu(es->s_inode_size); | |
3202 | sbi->s_first_ino = le32_to_cpu(es->s_first_ino); | |
617ba13b | 3203 | if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || |
1330593e | 3204 | (!is_power_of_2(sbi->s_inode_size)) || |
ac27a0ec | 3205 | (sbi->s_inode_size > blocksize)) { |
b31e1552 ES |
3206 | ext4_msg(sb, KERN_ERR, |
3207 | "unsupported inode size: %d", | |
2b2d6d01 | 3208 | sbi->s_inode_size); |
ac27a0ec DK |
3209 | goto failed_mount; |
3210 | } | |
ef7f3835 KS |
3211 | if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) |
3212 | sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2); | |
ac27a0ec | 3213 | } |
0b8e58a1 | 3214 | |
0d1ee42f AR |
3215 | sbi->s_desc_size = le16_to_cpu(es->s_desc_size); |
3216 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) { | |
8fadc143 | 3217 | if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || |
0d1ee42f | 3218 | sbi->s_desc_size > EXT4_MAX_DESC_SIZE || |
d8ea6cf8 | 3219 | !is_power_of_2(sbi->s_desc_size)) { |
b31e1552 ES |
3220 | ext4_msg(sb, KERN_ERR, |
3221 | "unsupported descriptor size %lu", | |
0d1ee42f AR |
3222 | sbi->s_desc_size); |
3223 | goto failed_mount; | |
3224 | } | |
3225 | } else | |
3226 | sbi->s_desc_size = EXT4_MIN_DESC_SIZE; | |
0b8e58a1 | 3227 | |
ac27a0ec | 3228 | sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); |
ac27a0ec | 3229 | sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); |
b47b6f38 | 3230 | if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0) |
617ba13b | 3231 | goto cantfind_ext4; |
0b8e58a1 | 3232 | |
617ba13b | 3233 | sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); |
ac27a0ec | 3234 | if (sbi->s_inodes_per_block == 0) |
617ba13b | 3235 | goto cantfind_ext4; |
ac27a0ec DK |
3236 | sbi->s_itb_per_group = sbi->s_inodes_per_group / |
3237 | sbi->s_inodes_per_block; | |
0d1ee42f | 3238 | sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); |
ac27a0ec DK |
3239 | sbi->s_sbh = bh; |
3240 | sbi->s_mount_state = le16_to_cpu(es->s_state); | |
e57aa839 FW |
3241 | sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); |
3242 | sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); | |
0b8e58a1 | 3243 | |
2b2d6d01 | 3244 | for (i = 0; i < 4; i++) |
ac27a0ec DK |
3245 | sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); |
3246 | sbi->s_def_hash_version = es->s_def_hash_version; | |
f99b2589 TT |
3247 | i = le32_to_cpu(es->s_flags); |
3248 | if (i & EXT2_FLAGS_UNSIGNED_HASH) | |
3249 | sbi->s_hash_unsigned = 3; | |
3250 | else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { | |
3251 | #ifdef __CHAR_UNSIGNED__ | |
3252 | es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); | |
3253 | sbi->s_hash_unsigned = 3; | |
3254 | #else | |
3255 | es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); | |
3256 | #endif | |
3257 | sb->s_dirt = 1; | |
3258 | } | |
ac27a0ec DK |
3259 | |
3260 | if (sbi->s_blocks_per_group > blocksize * 8) { | |
b31e1552 ES |
3261 | ext4_msg(sb, KERN_ERR, |
3262 | "#blocks per group too big: %lu", | |
2b2d6d01 | 3263 | sbi->s_blocks_per_group); |
ac27a0ec DK |
3264 | goto failed_mount; |
3265 | } | |
ac27a0ec | 3266 | if (sbi->s_inodes_per_group > blocksize * 8) { |
b31e1552 ES |
3267 | ext4_msg(sb, KERN_ERR, |
3268 | "#inodes per group too big: %lu", | |
2b2d6d01 | 3269 | sbi->s_inodes_per_group); |
ac27a0ec DK |
3270 | goto failed_mount; |
3271 | } | |
3272 | ||
bf43d84b ES |
3273 | /* |
3274 | * Test whether we have more sectors than will fit in sector_t, | |
3275 | * and whether the max offset is addressable by the page cache. | |
3276 | */ | |
5a9ae68a | 3277 | err = generic_check_addressable(sb->s_blocksize_bits, |
30ca22c7 | 3278 | ext4_blocks_count(es)); |
5a9ae68a | 3279 | if (err) { |
b31e1552 | 3280 | ext4_msg(sb, KERN_ERR, "filesystem" |
bf43d84b | 3281 | " too large to mount safely on this system"); |
ac27a0ec | 3282 | if (sizeof(sector_t) < 8) |
90c699a9 | 3283 | ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled"); |
5a9ae68a | 3284 | ret = err; |
ac27a0ec DK |
3285 | goto failed_mount; |
3286 | } | |
3287 | ||
617ba13b MC |
3288 | if (EXT4_BLOCKS_PER_GROUP(sb) == 0) |
3289 | goto cantfind_ext4; | |
e7c95593 | 3290 | |
0f2ddca6 FTN |
3291 | /* check blocks count against device size */ |
3292 | blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; | |
3293 | if (blocks_count && ext4_blocks_count(es) > blocks_count) { | |
b31e1552 ES |
3294 | ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " |
3295 | "exceeds size of device (%llu blocks)", | |
0f2ddca6 FTN |
3296 | ext4_blocks_count(es), blocks_count); |
3297 | goto failed_mount; | |
3298 | } | |
3299 | ||
0b8e58a1 AD |
3300 | /* |
3301 | * It makes no sense for the first data block to be beyond the end | |
3302 | * of the filesystem. | |
3303 | */ | |
3304 | if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { | |
b31e1552 ES |
3305 | ext4_msg(sb, KERN_WARNING, "bad geometry: first data" |
3306 | "block %u is beyond end of filesystem (%llu)", | |
3307 | le32_to_cpu(es->s_first_data_block), | |
3308 | ext4_blocks_count(es)); | |
e7c95593 ES |
3309 | goto failed_mount; |
3310 | } | |
bd81d8ee LV |
3311 | blocks_count = (ext4_blocks_count(es) - |
3312 | le32_to_cpu(es->s_first_data_block) + | |
3313 | EXT4_BLOCKS_PER_GROUP(sb) - 1); | |
3314 | do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); | |
4ec11028 | 3315 | if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { |
b31e1552 | 3316 | ext4_msg(sb, KERN_WARNING, "groups count too large: %u " |
4ec11028 | 3317 | "(block count %llu, first data block %u, " |
b31e1552 | 3318 | "blocks per group %lu)", sbi->s_groups_count, |
4ec11028 TT |
3319 | ext4_blocks_count(es), |
3320 | le32_to_cpu(es->s_first_data_block), | |
3321 | EXT4_BLOCKS_PER_GROUP(sb)); | |
3322 | goto failed_mount; | |
3323 | } | |
bd81d8ee | 3324 | sbi->s_groups_count = blocks_count; |
fb0a387d ES |
3325 | sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, |
3326 | (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); | |
617ba13b MC |
3327 | db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / |
3328 | EXT4_DESC_PER_BLOCK(sb); | |
2b2d6d01 | 3329 | sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *), |
ac27a0ec DK |
3330 | GFP_KERNEL); |
3331 | if (sbi->s_group_desc == NULL) { | |
b31e1552 | 3332 | ext4_msg(sb, KERN_ERR, "not enough memory"); |
ac27a0ec DK |
3333 | goto failed_mount; |
3334 | } | |
3335 | ||
3244fcb1 | 3336 | #ifdef CONFIG_PROC_FS |
9f6200bb TT |
3337 | if (ext4_proc_root) |
3338 | sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root); | |
3244fcb1 | 3339 | #endif |
240799cd | 3340 | |
705895b6 | 3341 | bgl_lock_init(sbi->s_blockgroup_lock); |
ac27a0ec DK |
3342 | |
3343 | for (i = 0; i < db_count; i++) { | |
70bbb3e0 | 3344 | block = descriptor_loc(sb, logical_sb_block, i); |
ac27a0ec DK |
3345 | sbi->s_group_desc[i] = sb_bread(sb, block); |
3346 | if (!sbi->s_group_desc[i]) { | |
b31e1552 ES |
3347 | ext4_msg(sb, KERN_ERR, |
3348 | "can't read group descriptor %d", i); | |
ac27a0ec DK |
3349 | db_count = i; |
3350 | goto failed_mount2; | |
3351 | } | |
3352 | } | |
bfff6873 | 3353 | if (!ext4_check_descriptors(sb, &first_not_zeroed)) { |
b31e1552 | 3354 | ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); |
ac27a0ec DK |
3355 | goto failed_mount2; |
3356 | } | |
772cb7c8 JS |
3357 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) |
3358 | if (!ext4_fill_flex_info(sb)) { | |
b31e1552 ES |
3359 | ext4_msg(sb, KERN_ERR, |
3360 | "unable to initialize " | |
3361 | "flex_bg meta info!"); | |
772cb7c8 JS |
3362 | goto failed_mount2; |
3363 | } | |
3364 | ||
ac27a0ec DK |
3365 | sbi->s_gdb_count = db_count; |
3366 | get_random_bytes(&sbi->s_next_generation, sizeof(u32)); | |
3367 | spin_lock_init(&sbi->s_next_gen_lock); | |
3368 | ||
ce7e010a TT |
3369 | err = percpu_counter_init(&sbi->s_freeblocks_counter, |
3370 | ext4_count_free_blocks(sb)); | |
3371 | if (!err) { | |
3372 | err = percpu_counter_init(&sbi->s_freeinodes_counter, | |
3373 | ext4_count_free_inodes(sb)); | |
3374 | } | |
3375 | if (!err) { | |
3376 | err = percpu_counter_init(&sbi->s_dirs_counter, | |
3377 | ext4_count_dirs(sb)); | |
3378 | } | |
3379 | if (!err) { | |
3380 | err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0); | |
3381 | } | |
3382 | if (err) { | |
3383 | ext4_msg(sb, KERN_ERR, "insufficient memory"); | |
3384 | goto failed_mount3; | |
3385 | } | |
3386 | ||
c9de560d | 3387 | sbi->s_stripe = ext4_get_stripe_size(sbi); |
55138e0b | 3388 | sbi->s_max_writeback_mb_bump = 128; |
c9de560d | 3389 | |
ac27a0ec DK |
3390 | /* |
3391 | * set up enough so that it can read an inode | |
3392 | */ | |
9ca92389 TT |
3393 | if (!test_opt(sb, NOLOAD) && |
3394 | EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) | |
3395 | sb->s_op = &ext4_sops; | |
3396 | else | |
3397 | sb->s_op = &ext4_nojournal_sops; | |
617ba13b MC |
3398 | sb->s_export_op = &ext4_export_ops; |
3399 | sb->s_xattr = ext4_xattr_handlers; | |
ac27a0ec | 3400 | #ifdef CONFIG_QUOTA |
617ba13b MC |
3401 | sb->s_qcop = &ext4_qctl_operations; |
3402 | sb->dq_op = &ext4_quota_operations; | |
ac27a0ec DK |
3403 | #endif |
3404 | INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ | |
3b9d4ed2 | 3405 | mutex_init(&sbi->s_orphan_lock); |
32ed5058 | 3406 | mutex_init(&sbi->s_resize_lock); |
ac27a0ec DK |
3407 | |
3408 | sb->s_root = NULL; | |
3409 | ||
3410 | needs_recovery = (es->s_last_orphan != 0 || | |
617ba13b MC |
3411 | EXT4_HAS_INCOMPAT_FEATURE(sb, |
3412 | EXT4_FEATURE_INCOMPAT_RECOVER)); | |
ac27a0ec DK |
3413 | |
3414 | /* | |
3415 | * The first inode we look at is the journal inode. Don't try | |
3416 | * root first: it may be modified in the journal! | |
3417 | */ | |
3418 | if (!test_opt(sb, NOLOAD) && | |
617ba13b MC |
3419 | EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { |
3420 | if (ext4_load_journal(sb, es, journal_devnum)) | |
ac27a0ec | 3421 | goto failed_mount3; |
0390131b FM |
3422 | } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) && |
3423 | EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { | |
b31e1552 ES |
3424 | ext4_msg(sb, KERN_ERR, "required journal recovery " |
3425 | "suppressed and not mounted read-only"); | |
744692dc | 3426 | goto failed_mount_wq; |
ac27a0ec | 3427 | } else { |
0390131b FM |
3428 | clear_opt(sbi->s_mount_opt, DATA_FLAGS); |
3429 | set_opt(sbi->s_mount_opt, WRITEBACK_DATA); | |
3430 | sbi->s_journal = NULL; | |
3431 | needs_recovery = 0; | |
3432 | goto no_journal; | |
ac27a0ec DK |
3433 | } |
3434 | ||
eb40a09c JS |
3435 | if (ext4_blocks_count(es) > 0xffffffffULL && |
3436 | !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, | |
3437 | JBD2_FEATURE_INCOMPAT_64BIT)) { | |
b31e1552 | 3438 | ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); |
744692dc | 3439 | goto failed_mount_wq; |
eb40a09c JS |
3440 | } |
3441 | ||
d4da6c9c LT |
3442 | if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { |
3443 | jbd2_journal_set_features(sbi->s_journal, | |
3444 | JBD2_FEATURE_COMPAT_CHECKSUM, 0, | |
818d276c | 3445 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); |
d4da6c9c LT |
3446 | } else if (test_opt(sb, JOURNAL_CHECKSUM)) { |
3447 | jbd2_journal_set_features(sbi->s_journal, | |
3448 | JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0); | |
818d276c GS |
3449 | jbd2_journal_clear_features(sbi->s_journal, 0, 0, |
3450 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); | |
d4da6c9c LT |
3451 | } else { |
3452 | jbd2_journal_clear_features(sbi->s_journal, | |
3453 | JBD2_FEATURE_COMPAT_CHECKSUM, 0, | |
3454 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); | |
3455 | } | |
818d276c | 3456 | |
ac27a0ec DK |
3457 | /* We have now updated the journal if required, so we can |
3458 | * validate the data journaling mode. */ | |
3459 | switch (test_opt(sb, DATA_FLAGS)) { | |
3460 | case 0: | |
3461 | /* No mode set, assume a default based on the journal | |
63f57933 AM |
3462 | * capabilities: ORDERED_DATA if the journal can |
3463 | * cope, else JOURNAL_DATA | |
3464 | */ | |
dab291af MC |
3465 | if (jbd2_journal_check_available_features |
3466 | (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) | |
ac27a0ec DK |
3467 | set_opt(sbi->s_mount_opt, ORDERED_DATA); |
3468 | else | |
3469 | set_opt(sbi->s_mount_opt, JOURNAL_DATA); | |
3470 | break; | |
3471 | ||
617ba13b MC |
3472 | case EXT4_MOUNT_ORDERED_DATA: |
3473 | case EXT4_MOUNT_WRITEBACK_DATA: | |
dab291af MC |
3474 | if (!jbd2_journal_check_available_features |
3475 | (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { | |
b31e1552 ES |
3476 | ext4_msg(sb, KERN_ERR, "Journal does not support " |
3477 | "requested data journaling mode"); | |
744692dc | 3478 | goto failed_mount_wq; |
ac27a0ec DK |
3479 | } |
3480 | default: | |
3481 | break; | |
3482 | } | |
b3881f74 | 3483 | set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); |
ac27a0ec | 3484 | |
ce7e010a TT |
3485 | /* |
3486 | * The journal may have updated the bg summary counts, so we | |
3487 | * need to update the global counters. | |
3488 | */ | |
3489 | percpu_counter_set(&sbi->s_freeblocks_counter, | |
3490 | ext4_count_free_blocks(sb)); | |
3491 | percpu_counter_set(&sbi->s_freeinodes_counter, | |
3492 | ext4_count_free_inodes(sb)); | |
3493 | percpu_counter_set(&sbi->s_dirs_counter, | |
3494 | ext4_count_dirs(sb)); | |
3495 | percpu_counter_set(&sbi->s_dirtyblocks_counter, 0); | |
206f7ab4 | 3496 | |
ce7e010a | 3497 | no_journal: |
4c0425ff MC |
3498 | EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten"); |
3499 | if (!EXT4_SB(sb)->dio_unwritten_wq) { | |
3500 | printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n"); | |
3501 | goto failed_mount_wq; | |
3502 | } | |
3503 | ||
ac27a0ec | 3504 | /* |
dab291af | 3505 | * The jbd2_journal_load will have done any necessary log recovery, |
ac27a0ec DK |
3506 | * so we can safely mount the rest of the filesystem now. |
3507 | */ | |
3508 | ||
1d1fe1ee DH |
3509 | root = ext4_iget(sb, EXT4_ROOT_INO); |
3510 | if (IS_ERR(root)) { | |
b31e1552 | 3511 | ext4_msg(sb, KERN_ERR, "get root inode failed"); |
1d1fe1ee | 3512 | ret = PTR_ERR(root); |
ac27a0ec DK |
3513 | goto failed_mount4; |
3514 | } | |
3515 | if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { | |
1d1fe1ee | 3516 | iput(root); |
b31e1552 | 3517 | ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); |
ac27a0ec DK |
3518 | goto failed_mount4; |
3519 | } | |
1d1fe1ee DH |
3520 | sb->s_root = d_alloc_root(root); |
3521 | if (!sb->s_root) { | |
b31e1552 | 3522 | ext4_msg(sb, KERN_ERR, "get root dentry failed"); |
1d1fe1ee DH |
3523 | iput(root); |
3524 | ret = -ENOMEM; | |
3525 | goto failed_mount4; | |
3526 | } | |
ac27a0ec | 3527 | |
2b2d6d01 | 3528 | ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY); |
ef7f3835 KS |
3529 | |
3530 | /* determine the minimum size of new large inodes, if present */ | |
3531 | if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { | |
3532 | sbi->s_want_extra_isize = sizeof(struct ext4_inode) - | |
3533 | EXT4_GOOD_OLD_INODE_SIZE; | |
3534 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, | |
3535 | EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) { | |
3536 | if (sbi->s_want_extra_isize < | |
3537 | le16_to_cpu(es->s_want_extra_isize)) | |
3538 | sbi->s_want_extra_isize = | |
3539 | le16_to_cpu(es->s_want_extra_isize); | |
3540 | if (sbi->s_want_extra_isize < | |
3541 | le16_to_cpu(es->s_min_extra_isize)) | |
3542 | sbi->s_want_extra_isize = | |
3543 | le16_to_cpu(es->s_min_extra_isize); | |
3544 | } | |
3545 | } | |
3546 | /* Check if enough inode space is available */ | |
3547 | if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize > | |
3548 | sbi->s_inode_size) { | |
3549 | sbi->s_want_extra_isize = sizeof(struct ext4_inode) - | |
3550 | EXT4_GOOD_OLD_INODE_SIZE; | |
b31e1552 ES |
3551 | ext4_msg(sb, KERN_INFO, "required extra inode space not" |
3552 | "available"); | |
ef7f3835 KS |
3553 | } |
3554 | ||
90576c0b TT |
3555 | if (test_opt(sb, DELALLOC) && |
3556 | (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) { | |
b31e1552 ES |
3557 | ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - " |
3558 | "requested data journaling mode"); | |
c2774d84 | 3559 | clear_opt(sbi->s_mount_opt, DELALLOC); |
90576c0b | 3560 | } |
744692dc JZ |
3561 | if (test_opt(sb, DIOREAD_NOLOCK)) { |
3562 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { | |
3563 | ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock " | |
3564 | "option - requested data journaling mode"); | |
3565 | clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK); | |
3566 | } | |
3567 | if (sb->s_blocksize < PAGE_SIZE) { | |
3568 | ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock " | |
3569 | "option - block size is too small"); | |
3570 | clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK); | |
3571 | } | |
3572 | } | |
c2774d84 | 3573 | |
6fd058f7 TT |
3574 | err = ext4_setup_system_zone(sb); |
3575 | if (err) { | |
b31e1552 | 3576 | ext4_msg(sb, KERN_ERR, "failed to initialize system " |
fbe845dd | 3577 | "zone (%d)", err); |
6fd058f7 TT |
3578 | goto failed_mount4; |
3579 | } | |
3580 | ||
c2774d84 AK |
3581 | ext4_ext_init(sb); |
3582 | err = ext4_mb_init(sb, needs_recovery); | |
3583 | if (err) { | |
421f91d2 | 3584 | ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", |
b31e1552 | 3585 | err); |
c2774d84 AK |
3586 | goto failed_mount4; |
3587 | } | |
3588 | ||
bfff6873 LC |
3589 | err = ext4_register_li_request(sb, first_not_zeroed); |
3590 | if (err) | |
3591 | goto failed_mount4; | |
3592 | ||
3197ebdb TT |
3593 | sbi->s_kobj.kset = ext4_kset; |
3594 | init_completion(&sbi->s_kobj_unregister); | |
3595 | err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL, | |
3596 | "%s", sb->s_id); | |
3597 | if (err) { | |
3598 | ext4_mb_release(sb); | |
3599 | ext4_ext_release(sb); | |
3600 | goto failed_mount4; | |
3601 | }; | |
3602 | ||
617ba13b MC |
3603 | EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; |
3604 | ext4_orphan_cleanup(sb, es); | |
3605 | EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; | |
0390131b | 3606 | if (needs_recovery) { |
b31e1552 | 3607 | ext4_msg(sb, KERN_INFO, "recovery complete"); |
0390131b FM |
3608 | ext4_mark_recovery_complete(sb, es); |
3609 | } | |
3610 | if (EXT4_SB(sb)->s_journal) { | |
3611 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) | |
3612 | descr = " journalled data mode"; | |
3613 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) | |
3614 | descr = " ordered data mode"; | |
3615 | else | |
3616 | descr = " writeback data mode"; | |
3617 | } else | |
3618 | descr = "out journal"; | |
3619 | ||
d4c402d9 | 3620 | ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " |
8b67f04a TT |
3621 | "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts, |
3622 | *sbi->s_es->s_mount_opts ? "; " : "", orig_data); | |
ac27a0ec | 3623 | |
66e61a9e TT |
3624 | init_timer(&sbi->s_err_report); |
3625 | sbi->s_err_report.function = print_daily_error_info; | |
3626 | sbi->s_err_report.data = (unsigned long) sb; | |
3627 | if (es->s_error_count) | |
3628 | mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ | |
ac27a0ec | 3629 | |
d4c402d9 | 3630 | kfree(orig_data); |
ac27a0ec DK |
3631 | return 0; |
3632 | ||
617ba13b | 3633 | cantfind_ext4: |
ac27a0ec | 3634 | if (!silent) |
b31e1552 | 3635 | ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); |
ac27a0ec DK |
3636 | goto failed_mount; |
3637 | ||
3638 | failed_mount4: | |
b31e1552 | 3639 | ext4_msg(sb, KERN_ERR, "mount failed"); |
4c0425ff MC |
3640 | destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq); |
3641 | failed_mount_wq: | |
6fd058f7 | 3642 | ext4_release_system_zone(sb); |
0390131b FM |
3643 | if (sbi->s_journal) { |
3644 | jbd2_journal_destroy(sbi->s_journal); | |
3645 | sbi->s_journal = NULL; | |
3646 | } | |
ac27a0ec | 3647 | failed_mount3: |
c5ca7c76 TT |
3648 | if (sbi->s_flex_groups) { |
3649 | if (is_vmalloc_addr(sbi->s_flex_groups)) | |
3650 | vfree(sbi->s_flex_groups); | |
3651 | else | |
3652 | kfree(sbi->s_flex_groups); | |
3653 | } | |
ce7e010a TT |
3654 | percpu_counter_destroy(&sbi->s_freeblocks_counter); |
3655 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
3656 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
3657 | percpu_counter_destroy(&sbi->s_dirtyblocks_counter); | |
ac27a0ec DK |
3658 | failed_mount2: |
3659 | for (i = 0; i < db_count; i++) | |
3660 | brelse(sbi->s_group_desc[i]); | |
3661 | kfree(sbi->s_group_desc); | |
3662 | failed_mount: | |
240799cd | 3663 | if (sbi->s_proc) { |
9f6200bb | 3664 | remove_proc_entry(sb->s_id, ext4_proc_root); |
240799cd | 3665 | } |
ac27a0ec DK |
3666 | #ifdef CONFIG_QUOTA |
3667 | for (i = 0; i < MAXQUOTAS; i++) | |
3668 | kfree(sbi->s_qf_names[i]); | |
3669 | #endif | |
617ba13b | 3670 | ext4_blkdev_remove(sbi); |
ac27a0ec DK |
3671 | brelse(bh); |
3672 | out_fail: | |
3673 | sb->s_fs_info = NULL; | |
f6830165 | 3674 | kfree(sbi->s_blockgroup_lock); |
ac27a0ec | 3675 | kfree(sbi); |
dcc7dae3 | 3676 | out_free_orig: |
d4c402d9 | 3677 | kfree(orig_data); |
1d1fe1ee | 3678 | return ret; |
ac27a0ec DK |
3679 | } |
3680 | ||
3681 | /* | |
3682 | * Setup any per-fs journal parameters now. We'll do this both on | |
3683 | * initial mount, once the journal has been initialised but before we've | |
3684 | * done any recovery; and again on any subsequent remount. | |
3685 | */ | |
617ba13b | 3686 | static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) |
ac27a0ec | 3687 | { |
617ba13b | 3688 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 3689 | |
30773840 TT |
3690 | journal->j_commit_interval = sbi->s_commit_interval; |
3691 | journal->j_min_batch_time = sbi->s_min_batch_time; | |
3692 | journal->j_max_batch_time = sbi->s_max_batch_time; | |
ac27a0ec | 3693 | |
a931da6a | 3694 | write_lock(&journal->j_state_lock); |
ac27a0ec | 3695 | if (test_opt(sb, BARRIER)) |
dab291af | 3696 | journal->j_flags |= JBD2_BARRIER; |
ac27a0ec | 3697 | else |
dab291af | 3698 | journal->j_flags &= ~JBD2_BARRIER; |
5bf5683a HK |
3699 | if (test_opt(sb, DATA_ERR_ABORT)) |
3700 | journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; | |
3701 | else | |
3702 | journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; | |
a931da6a | 3703 | write_unlock(&journal->j_state_lock); |
ac27a0ec DK |
3704 | } |
3705 | ||
617ba13b | 3706 | static journal_t *ext4_get_journal(struct super_block *sb, |
ac27a0ec DK |
3707 | unsigned int journal_inum) |
3708 | { | |
3709 | struct inode *journal_inode; | |
3710 | journal_t *journal; | |
3711 | ||
0390131b FM |
3712 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
3713 | ||
ac27a0ec DK |
3714 | /* First, test for the existence of a valid inode on disk. Bad |
3715 | * things happen if we iget() an unused inode, as the subsequent | |
3716 | * iput() will try to delete it. */ | |
3717 | ||
1d1fe1ee DH |
3718 | journal_inode = ext4_iget(sb, journal_inum); |
3719 | if (IS_ERR(journal_inode)) { | |
b31e1552 | 3720 | ext4_msg(sb, KERN_ERR, "no journal found"); |
ac27a0ec DK |
3721 | return NULL; |
3722 | } | |
3723 | if (!journal_inode->i_nlink) { | |
3724 | make_bad_inode(journal_inode); | |
3725 | iput(journal_inode); | |
b31e1552 | 3726 | ext4_msg(sb, KERN_ERR, "journal inode is deleted"); |
ac27a0ec DK |
3727 | return NULL; |
3728 | } | |
3729 | ||
e5f8eab8 | 3730 | jbd_debug(2, "Journal inode found at %p: %lld bytes\n", |
ac27a0ec | 3731 | journal_inode, journal_inode->i_size); |
1d1fe1ee | 3732 | if (!S_ISREG(journal_inode->i_mode)) { |
b31e1552 | 3733 | ext4_msg(sb, KERN_ERR, "invalid journal inode"); |
ac27a0ec DK |
3734 | iput(journal_inode); |
3735 | return NULL; | |
3736 | } | |
3737 | ||
dab291af | 3738 | journal = jbd2_journal_init_inode(journal_inode); |
ac27a0ec | 3739 | if (!journal) { |
b31e1552 | 3740 | ext4_msg(sb, KERN_ERR, "Could not load journal inode"); |
ac27a0ec DK |
3741 | iput(journal_inode); |
3742 | return NULL; | |
3743 | } | |
3744 | journal->j_private = sb; | |
617ba13b | 3745 | ext4_init_journal_params(sb, journal); |
ac27a0ec DK |
3746 | return journal; |
3747 | } | |
3748 | ||
617ba13b | 3749 | static journal_t *ext4_get_dev_journal(struct super_block *sb, |
ac27a0ec DK |
3750 | dev_t j_dev) |
3751 | { | |
2b2d6d01 | 3752 | struct buffer_head *bh; |
ac27a0ec | 3753 | journal_t *journal; |
617ba13b MC |
3754 | ext4_fsblk_t start; |
3755 | ext4_fsblk_t len; | |
ac27a0ec | 3756 | int hblock, blocksize; |
617ba13b | 3757 | ext4_fsblk_t sb_block; |
ac27a0ec | 3758 | unsigned long offset; |
2b2d6d01 | 3759 | struct ext4_super_block *es; |
ac27a0ec DK |
3760 | struct block_device *bdev; |
3761 | ||
0390131b FM |
3762 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
3763 | ||
b31e1552 | 3764 | bdev = ext4_blkdev_get(j_dev, sb); |
ac27a0ec DK |
3765 | if (bdev == NULL) |
3766 | return NULL; | |
3767 | ||
3768 | if (bd_claim(bdev, sb)) { | |
b31e1552 ES |
3769 | ext4_msg(sb, KERN_ERR, |
3770 | "failed to claim external journal device"); | |
9a1c3542 | 3771 | blkdev_put(bdev, FMODE_READ|FMODE_WRITE); |
ac27a0ec DK |
3772 | return NULL; |
3773 | } | |
3774 | ||
3775 | blocksize = sb->s_blocksize; | |
e1defc4f | 3776 | hblock = bdev_logical_block_size(bdev); |
ac27a0ec | 3777 | if (blocksize < hblock) { |
b31e1552 ES |
3778 | ext4_msg(sb, KERN_ERR, |
3779 | "blocksize too small for journal device"); | |
ac27a0ec DK |
3780 | goto out_bdev; |
3781 | } | |
3782 | ||
617ba13b MC |
3783 | sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; |
3784 | offset = EXT4_MIN_BLOCK_SIZE % blocksize; | |
ac27a0ec DK |
3785 | set_blocksize(bdev, blocksize); |
3786 | if (!(bh = __bread(bdev, sb_block, blocksize))) { | |
b31e1552 ES |
3787 | ext4_msg(sb, KERN_ERR, "couldn't read superblock of " |
3788 | "external journal"); | |
ac27a0ec DK |
3789 | goto out_bdev; |
3790 | } | |
3791 | ||
617ba13b MC |
3792 | es = (struct ext4_super_block *) (((char *)bh->b_data) + offset); |
3793 | if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || | |
ac27a0ec | 3794 | !(le32_to_cpu(es->s_feature_incompat) & |
617ba13b | 3795 | EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { |
b31e1552 ES |
3796 | ext4_msg(sb, KERN_ERR, "external journal has " |
3797 | "bad superblock"); | |
ac27a0ec DK |
3798 | brelse(bh); |
3799 | goto out_bdev; | |
3800 | } | |
3801 | ||
617ba13b | 3802 | if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { |
b31e1552 | 3803 | ext4_msg(sb, KERN_ERR, "journal UUID does not match"); |
ac27a0ec DK |
3804 | brelse(bh); |
3805 | goto out_bdev; | |
3806 | } | |
3807 | ||
bd81d8ee | 3808 | len = ext4_blocks_count(es); |
ac27a0ec DK |
3809 | start = sb_block + 1; |
3810 | brelse(bh); /* we're done with the superblock */ | |
3811 | ||
dab291af | 3812 | journal = jbd2_journal_init_dev(bdev, sb->s_bdev, |
ac27a0ec DK |
3813 | start, len, blocksize); |
3814 | if (!journal) { | |
b31e1552 | 3815 | ext4_msg(sb, KERN_ERR, "failed to create device journal"); |
ac27a0ec DK |
3816 | goto out_bdev; |
3817 | } | |
3818 | journal->j_private = sb; | |
3819 | ll_rw_block(READ, 1, &journal->j_sb_buffer); | |
3820 | wait_on_buffer(journal->j_sb_buffer); | |
3821 | if (!buffer_uptodate(journal->j_sb_buffer)) { | |
b31e1552 | 3822 | ext4_msg(sb, KERN_ERR, "I/O error on journal device"); |
ac27a0ec DK |
3823 | goto out_journal; |
3824 | } | |
3825 | if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { | |
b31e1552 ES |
3826 | ext4_msg(sb, KERN_ERR, "External journal has more than one " |
3827 | "user (unsupported) - %d", | |
ac27a0ec DK |
3828 | be32_to_cpu(journal->j_superblock->s_nr_users)); |
3829 | goto out_journal; | |
3830 | } | |
617ba13b MC |
3831 | EXT4_SB(sb)->journal_bdev = bdev; |
3832 | ext4_init_journal_params(sb, journal); | |
ac27a0ec | 3833 | return journal; |
0b8e58a1 | 3834 | |
ac27a0ec | 3835 | out_journal: |
dab291af | 3836 | jbd2_journal_destroy(journal); |
ac27a0ec | 3837 | out_bdev: |
617ba13b | 3838 | ext4_blkdev_put(bdev); |
ac27a0ec DK |
3839 | return NULL; |
3840 | } | |
3841 | ||
617ba13b MC |
3842 | static int ext4_load_journal(struct super_block *sb, |
3843 | struct ext4_super_block *es, | |
ac27a0ec DK |
3844 | unsigned long journal_devnum) |
3845 | { | |
3846 | journal_t *journal; | |
3847 | unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); | |
3848 | dev_t journal_dev; | |
3849 | int err = 0; | |
3850 | int really_read_only; | |
3851 | ||
0390131b FM |
3852 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
3853 | ||
ac27a0ec DK |
3854 | if (journal_devnum && |
3855 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { | |
b31e1552 ES |
3856 | ext4_msg(sb, KERN_INFO, "external journal device major/minor " |
3857 | "numbers have changed"); | |
ac27a0ec DK |
3858 | journal_dev = new_decode_dev(journal_devnum); |
3859 | } else | |
3860 | journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); | |
3861 | ||
3862 | really_read_only = bdev_read_only(sb->s_bdev); | |
3863 | ||
3864 | /* | |
3865 | * Are we loading a blank journal or performing recovery after a | |
3866 | * crash? For recovery, we need to check in advance whether we | |
3867 | * can get read-write access to the device. | |
3868 | */ | |
617ba13b | 3869 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { |
ac27a0ec | 3870 | if (sb->s_flags & MS_RDONLY) { |
b31e1552 ES |
3871 | ext4_msg(sb, KERN_INFO, "INFO: recovery " |
3872 | "required on readonly filesystem"); | |
ac27a0ec | 3873 | if (really_read_only) { |
b31e1552 ES |
3874 | ext4_msg(sb, KERN_ERR, "write access " |
3875 | "unavailable, cannot proceed"); | |
ac27a0ec DK |
3876 | return -EROFS; |
3877 | } | |
b31e1552 ES |
3878 | ext4_msg(sb, KERN_INFO, "write access will " |
3879 | "be enabled during recovery"); | |
ac27a0ec DK |
3880 | } |
3881 | } | |
3882 | ||
3883 | if (journal_inum && journal_dev) { | |
b31e1552 ES |
3884 | ext4_msg(sb, KERN_ERR, "filesystem has both journal " |
3885 | "and inode journals!"); | |
ac27a0ec DK |
3886 | return -EINVAL; |
3887 | } | |
3888 | ||
3889 | if (journal_inum) { | |
617ba13b | 3890 | if (!(journal = ext4_get_journal(sb, journal_inum))) |
ac27a0ec DK |
3891 | return -EINVAL; |
3892 | } else { | |
617ba13b | 3893 | if (!(journal = ext4_get_dev_journal(sb, journal_dev))) |
ac27a0ec DK |
3894 | return -EINVAL; |
3895 | } | |
3896 | ||
90576c0b | 3897 | if (!(journal->j_flags & JBD2_BARRIER)) |
b31e1552 | 3898 | ext4_msg(sb, KERN_INFO, "barriers disabled"); |
4776004f | 3899 | |
ac27a0ec | 3900 | if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) { |
dab291af | 3901 | err = jbd2_journal_update_format(journal); |
ac27a0ec | 3902 | if (err) { |
b31e1552 | 3903 | ext4_msg(sb, KERN_ERR, "error updating journal"); |
dab291af | 3904 | jbd2_journal_destroy(journal); |
ac27a0ec DK |
3905 | return err; |
3906 | } | |
3907 | } | |
3908 | ||
617ba13b | 3909 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) |
dab291af | 3910 | err = jbd2_journal_wipe(journal, !really_read_only); |
1c13d5c0 TT |
3911 | if (!err) { |
3912 | char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); | |
3913 | if (save) | |
3914 | memcpy(save, ((char *) es) + | |
3915 | EXT4_S_ERR_START, EXT4_S_ERR_LEN); | |
dab291af | 3916 | err = jbd2_journal_load(journal); |
1c13d5c0 TT |
3917 | if (save) |
3918 | memcpy(((char *) es) + EXT4_S_ERR_START, | |
3919 | save, EXT4_S_ERR_LEN); | |
3920 | kfree(save); | |
3921 | } | |
ac27a0ec DK |
3922 | |
3923 | if (err) { | |
b31e1552 | 3924 | ext4_msg(sb, KERN_ERR, "error loading journal"); |
dab291af | 3925 | jbd2_journal_destroy(journal); |
ac27a0ec DK |
3926 | return err; |
3927 | } | |
3928 | ||
617ba13b MC |
3929 | EXT4_SB(sb)->s_journal = journal; |
3930 | ext4_clear_journal_err(sb, es); | |
ac27a0ec | 3931 | |
c41303ce | 3932 | if (!really_read_only && journal_devnum && |
ac27a0ec DK |
3933 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { |
3934 | es->s_journal_dev = cpu_to_le32(journal_devnum); | |
ac27a0ec DK |
3935 | |
3936 | /* Make sure we flush the recovery flag to disk. */ | |
e2d67052 | 3937 | ext4_commit_super(sb, 1); |
ac27a0ec DK |
3938 | } |
3939 | ||
3940 | return 0; | |
3941 | } | |
3942 | ||
e2d67052 | 3943 | static int ext4_commit_super(struct super_block *sb, int sync) |
ac27a0ec | 3944 | { |
e2d67052 | 3945 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
617ba13b | 3946 | struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; |
c4be0c1d | 3947 | int error = 0; |
ac27a0ec DK |
3948 | |
3949 | if (!sbh) | |
c4be0c1d | 3950 | return error; |
914258bf TT |
3951 | if (buffer_write_io_error(sbh)) { |
3952 | /* | |
3953 | * Oh, dear. A previous attempt to write the | |
3954 | * superblock failed. This could happen because the | |
3955 | * USB device was yanked out. Or it could happen to | |
3956 | * be a transient write error and maybe the block will | |
3957 | * be remapped. Nothing we can do but to retry the | |
3958 | * write and hope for the best. | |
3959 | */ | |
b31e1552 ES |
3960 | ext4_msg(sb, KERN_ERR, "previous I/O error to " |
3961 | "superblock detected"); | |
914258bf TT |
3962 | clear_buffer_write_io_error(sbh); |
3963 | set_buffer_uptodate(sbh); | |
3964 | } | |
71290b36 TT |
3965 | /* |
3966 | * If the file system is mounted read-only, don't update the | |
3967 | * superblock write time. This avoids updating the superblock | |
3968 | * write time when we are mounting the root file system | |
3969 | * read/only but we need to replay the journal; at that point, | |
3970 | * for people who are east of GMT and who make their clock | |
3971 | * tick in localtime for Windows bug-for-bug compatibility, | |
3972 | * the clock is set in the future, and this will cause e2fsck | |
3973 | * to complain and force a full file system check. | |
3974 | */ | |
3975 | if (!(sb->s_flags & MS_RDONLY)) | |
3976 | es->s_wtime = cpu_to_le32(get_seconds()); | |
f613dfcb TT |
3977 | if (sb->s_bdev->bd_part) |
3978 | es->s_kbytes_written = | |
3979 | cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + | |
afc32f7e TT |
3980 | ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - |
3981 | EXT4_SB(sb)->s_sectors_written_start) >> 1)); | |
f613dfcb TT |
3982 | else |
3983 | es->s_kbytes_written = | |
3984 | cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); | |
ce7e010a TT |
3985 | ext4_free_blocks_count_set(es, percpu_counter_sum_positive( |
3986 | &EXT4_SB(sb)->s_freeblocks_counter)); | |
3987 | es->s_free_inodes_count = | |
3988 | cpu_to_le32(percpu_counter_sum_positive( | |
3989 | &EXT4_SB(sb)->s_freeinodes_counter)); | |
7234ab2a | 3990 | sb->s_dirt = 0; |
ac27a0ec DK |
3991 | BUFFER_TRACE(sbh, "marking dirty"); |
3992 | mark_buffer_dirty(sbh); | |
914258bf | 3993 | if (sync) { |
c4be0c1d TS |
3994 | error = sync_dirty_buffer(sbh); |
3995 | if (error) | |
3996 | return error; | |
3997 | ||
3998 | error = buffer_write_io_error(sbh); | |
3999 | if (error) { | |
b31e1552 ES |
4000 | ext4_msg(sb, KERN_ERR, "I/O error while writing " |
4001 | "superblock"); | |
914258bf TT |
4002 | clear_buffer_write_io_error(sbh); |
4003 | set_buffer_uptodate(sbh); | |
4004 | } | |
4005 | } | |
c4be0c1d | 4006 | return error; |
ac27a0ec DK |
4007 | } |
4008 | ||
ac27a0ec DK |
4009 | /* |
4010 | * Have we just finished recovery? If so, and if we are mounting (or | |
4011 | * remounting) the filesystem readonly, then we will end up with a | |
4012 | * consistent fs on disk. Record that fact. | |
4013 | */ | |
2b2d6d01 TT |
4014 | static void ext4_mark_recovery_complete(struct super_block *sb, |
4015 | struct ext4_super_block *es) | |
ac27a0ec | 4016 | { |
617ba13b | 4017 | journal_t *journal = EXT4_SB(sb)->s_journal; |
ac27a0ec | 4018 | |
0390131b FM |
4019 | if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { |
4020 | BUG_ON(journal != NULL); | |
4021 | return; | |
4022 | } | |
dab291af | 4023 | jbd2_journal_lock_updates(journal); |
7ffe1ea8 HK |
4024 | if (jbd2_journal_flush(journal) < 0) |
4025 | goto out; | |
4026 | ||
617ba13b | 4027 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) && |
ac27a0ec | 4028 | sb->s_flags & MS_RDONLY) { |
617ba13b | 4029 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); |
e2d67052 | 4030 | ext4_commit_super(sb, 1); |
ac27a0ec | 4031 | } |
7ffe1ea8 HK |
4032 | |
4033 | out: | |
dab291af | 4034 | jbd2_journal_unlock_updates(journal); |
ac27a0ec DK |
4035 | } |
4036 | ||
4037 | /* | |
4038 | * If we are mounting (or read-write remounting) a filesystem whose journal | |
4039 | * has recorded an error from a previous lifetime, move that error to the | |
4040 | * main filesystem now. | |
4041 | */ | |
2b2d6d01 TT |
4042 | static void ext4_clear_journal_err(struct super_block *sb, |
4043 | struct ext4_super_block *es) | |
ac27a0ec DK |
4044 | { |
4045 | journal_t *journal; | |
4046 | int j_errno; | |
4047 | const char *errstr; | |
4048 | ||
0390131b FM |
4049 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
4050 | ||
617ba13b | 4051 | journal = EXT4_SB(sb)->s_journal; |
ac27a0ec DK |
4052 | |
4053 | /* | |
4054 | * Now check for any error status which may have been recorded in the | |
617ba13b | 4055 | * journal by a prior ext4_error() or ext4_abort() |
ac27a0ec DK |
4056 | */ |
4057 | ||
dab291af | 4058 | j_errno = jbd2_journal_errno(journal); |
ac27a0ec DK |
4059 | if (j_errno) { |
4060 | char nbuf[16]; | |
4061 | ||
617ba13b | 4062 | errstr = ext4_decode_error(sb, j_errno, nbuf); |
12062ddd | 4063 | ext4_warning(sb, "Filesystem error recorded " |
ac27a0ec | 4064 | "from previous mount: %s", errstr); |
12062ddd | 4065 | ext4_warning(sb, "Marking fs in need of filesystem check."); |
ac27a0ec | 4066 | |
617ba13b MC |
4067 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; |
4068 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
e2d67052 | 4069 | ext4_commit_super(sb, 1); |
ac27a0ec | 4070 | |
dab291af | 4071 | jbd2_journal_clear_err(journal); |
ac27a0ec DK |
4072 | } |
4073 | } | |
4074 | ||
4075 | /* | |
4076 | * Force the running and committing transactions to commit, | |
4077 | * and wait on the commit. | |
4078 | */ | |
617ba13b | 4079 | int ext4_force_commit(struct super_block *sb) |
ac27a0ec DK |
4080 | { |
4081 | journal_t *journal; | |
0390131b | 4082 | int ret = 0; |
ac27a0ec DK |
4083 | |
4084 | if (sb->s_flags & MS_RDONLY) | |
4085 | return 0; | |
4086 | ||
617ba13b | 4087 | journal = EXT4_SB(sb)->s_journal; |
6b0310fb | 4088 | if (journal) { |
437f88cc | 4089 | vfs_check_frozen(sb, SB_FREEZE_TRANS); |
0390131b | 4090 | ret = ext4_journal_force_commit(journal); |
6b0310fb | 4091 | } |
0390131b | 4092 | |
ac27a0ec DK |
4093 | return ret; |
4094 | } | |
4095 | ||
2b2d6d01 | 4096 | static void ext4_write_super(struct super_block *sb) |
ac27a0ec | 4097 | { |
ebc1ac16 | 4098 | lock_super(sb); |
9ca92389 | 4099 | ext4_commit_super(sb, 1); |
ebc1ac16 | 4100 | unlock_super(sb); |
ac27a0ec DK |
4101 | } |
4102 | ||
617ba13b | 4103 | static int ext4_sync_fs(struct super_block *sb, int wait) |
ac27a0ec | 4104 | { |
14ce0cb4 | 4105 | int ret = 0; |
9eddacf9 | 4106 | tid_t target; |
8d5d02e6 | 4107 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 4108 | |
9bffad1e | 4109 | trace_ext4_sync_fs(sb, wait); |
8d5d02e6 MC |
4110 | flush_workqueue(sbi->dio_unwritten_wq); |
4111 | if (jbd2_journal_start_commit(sbi->s_journal, &target)) { | |
9ca92389 | 4112 | if (wait) |
8d5d02e6 | 4113 | jbd2_log_wait_commit(sbi->s_journal, target); |
0390131b | 4114 | } |
14ce0cb4 | 4115 | return ret; |
ac27a0ec DK |
4116 | } |
4117 | ||
4118 | /* | |
4119 | * LVM calls this function before a (read-only) snapshot is created. This | |
4120 | * gives us a chance to flush the journal completely and mark the fs clean. | |
4121 | */ | |
c4be0c1d | 4122 | static int ext4_freeze(struct super_block *sb) |
ac27a0ec | 4123 | { |
c4be0c1d TS |
4124 | int error = 0; |
4125 | journal_t *journal; | |
ac27a0ec | 4126 | |
9ca92389 TT |
4127 | if (sb->s_flags & MS_RDONLY) |
4128 | return 0; | |
ac27a0ec | 4129 | |
9ca92389 | 4130 | journal = EXT4_SB(sb)->s_journal; |
7ffe1ea8 | 4131 | |
9ca92389 TT |
4132 | /* Now we set up the journal barrier. */ |
4133 | jbd2_journal_lock_updates(journal); | |
ac27a0ec | 4134 | |
9ca92389 TT |
4135 | /* |
4136 | * Don't clear the needs_recovery flag if we failed to flush | |
4137 | * the journal. | |
4138 | */ | |
4139 | error = jbd2_journal_flush(journal); | |
6b0310fb ES |
4140 | if (error < 0) |
4141 | goto out; | |
9ca92389 TT |
4142 | |
4143 | /* Journal blocked and flushed, clear needs_recovery flag. */ | |
4144 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
4145 | error = ext4_commit_super(sb, 1); | |
6b0310fb ES |
4146 | out: |
4147 | /* we rely on s_frozen to stop further updates */ | |
4148 | jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); | |
4149 | return error; | |
ac27a0ec DK |
4150 | } |
4151 | ||
4152 | /* | |
4153 | * Called by LVM after the snapshot is done. We need to reset the RECOVER | |
4154 | * flag here, even though the filesystem is not technically dirty yet. | |
4155 | */ | |
c4be0c1d | 4156 | static int ext4_unfreeze(struct super_block *sb) |
ac27a0ec | 4157 | { |
9ca92389 TT |
4158 | if (sb->s_flags & MS_RDONLY) |
4159 | return 0; | |
4160 | ||
4161 | lock_super(sb); | |
4162 | /* Reset the needs_recovery flag before the fs is unlocked. */ | |
4163 | EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
4164 | ext4_commit_super(sb, 1); | |
4165 | unlock_super(sb); | |
c4be0c1d | 4166 | return 0; |
ac27a0ec DK |
4167 | } |
4168 | ||
2b2d6d01 | 4169 | static int ext4_remount(struct super_block *sb, int *flags, char *data) |
ac27a0ec | 4170 | { |
2b2d6d01 | 4171 | struct ext4_super_block *es; |
617ba13b MC |
4172 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
4173 | ext4_fsblk_t n_blocks_count = 0; | |
ac27a0ec | 4174 | unsigned long old_sb_flags; |
617ba13b | 4175 | struct ext4_mount_options old_opts; |
c79d967d | 4176 | int enable_quota = 0; |
8a266467 | 4177 | ext4_group_t g; |
b3881f74 | 4178 | unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; |
ac27a0ec DK |
4179 | int err; |
4180 | #ifdef CONFIG_QUOTA | |
4181 | int i; | |
4182 | #endif | |
d4c402d9 | 4183 | char *orig_data = kstrdup(data, GFP_KERNEL); |
ac27a0ec DK |
4184 | |
4185 | /* Store the original options */ | |
bbd6851a | 4186 | lock_super(sb); |
ac27a0ec DK |
4187 | old_sb_flags = sb->s_flags; |
4188 | old_opts.s_mount_opt = sbi->s_mount_opt; | |
4189 | old_opts.s_resuid = sbi->s_resuid; | |
4190 | old_opts.s_resgid = sbi->s_resgid; | |
4191 | old_opts.s_commit_interval = sbi->s_commit_interval; | |
30773840 TT |
4192 | old_opts.s_min_batch_time = sbi->s_min_batch_time; |
4193 | old_opts.s_max_batch_time = sbi->s_max_batch_time; | |
ac27a0ec DK |
4194 | #ifdef CONFIG_QUOTA |
4195 | old_opts.s_jquota_fmt = sbi->s_jquota_fmt; | |
4196 | for (i = 0; i < MAXQUOTAS; i++) | |
4197 | old_opts.s_qf_names[i] = sbi->s_qf_names[i]; | |
4198 | #endif | |
b3881f74 TT |
4199 | if (sbi->s_journal && sbi->s_journal->j_task->io_context) |
4200 | journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; | |
ac27a0ec DK |
4201 | |
4202 | /* | |
4203 | * Allow the "check" option to be passed as a remount option. | |
4204 | */ | |
b3881f74 TT |
4205 | if (!parse_options(data, sb, NULL, &journal_ioprio, |
4206 | &n_blocks_count, 1)) { | |
ac27a0ec DK |
4207 | err = -EINVAL; |
4208 | goto restore_opts; | |
4209 | } | |
4210 | ||
4ab2f15b | 4211 | if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) |
c67d859e | 4212 | ext4_abort(sb, "Abort forced by user"); |
ac27a0ec DK |
4213 | |
4214 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
482a7425 | 4215 | (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); |
ac27a0ec DK |
4216 | |
4217 | es = sbi->s_es; | |
4218 | ||
b3881f74 | 4219 | if (sbi->s_journal) { |
0390131b | 4220 | ext4_init_journal_params(sb, sbi->s_journal); |
b3881f74 TT |
4221 | set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); |
4222 | } | |
ac27a0ec DK |
4223 | |
4224 | if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) || | |
bd81d8ee | 4225 | n_blocks_count > ext4_blocks_count(es)) { |
4ab2f15b | 4226 | if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { |
ac27a0ec DK |
4227 | err = -EROFS; |
4228 | goto restore_opts; | |
4229 | } | |
4230 | ||
4231 | if (*flags & MS_RDONLY) { | |
0f0dd62f CH |
4232 | err = dquot_suspend(sb, -1); |
4233 | if (err < 0) | |
c79d967d | 4234 | goto restore_opts; |
c79d967d | 4235 | |
ac27a0ec DK |
4236 | /* |
4237 | * First of all, the unconditional stuff we have to do | |
4238 | * to disable replay of the journal when we next remount | |
4239 | */ | |
4240 | sb->s_flags |= MS_RDONLY; | |
4241 | ||
4242 | /* | |
4243 | * OK, test if we are remounting a valid rw partition | |
4244 | * readonly, and if so set the rdonly flag and then | |
4245 | * mark the partition as valid again. | |
4246 | */ | |
617ba13b MC |
4247 | if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && |
4248 | (sbi->s_mount_state & EXT4_VALID_FS)) | |
ac27a0ec DK |
4249 | es->s_state = cpu_to_le16(sbi->s_mount_state); |
4250 | ||
a63c9eb2 | 4251 | if (sbi->s_journal) |
0390131b | 4252 | ext4_mark_recovery_complete(sb, es); |
ac27a0ec | 4253 | } else { |
a13fb1a4 ES |
4254 | /* Make sure we can mount this feature set readwrite */ |
4255 | if (!ext4_feature_set_ok(sb, 0)) { | |
ac27a0ec DK |
4256 | err = -EROFS; |
4257 | goto restore_opts; | |
4258 | } | |
8a266467 TT |
4259 | /* |
4260 | * Make sure the group descriptor checksums | |
0b8e58a1 | 4261 | * are sane. If they aren't, refuse to remount r/w. |
8a266467 TT |
4262 | */ |
4263 | for (g = 0; g < sbi->s_groups_count; g++) { | |
4264 | struct ext4_group_desc *gdp = | |
4265 | ext4_get_group_desc(sb, g, NULL); | |
4266 | ||
4267 | if (!ext4_group_desc_csum_verify(sbi, g, gdp)) { | |
b31e1552 ES |
4268 | ext4_msg(sb, KERN_ERR, |
4269 | "ext4_remount: Checksum for group %u failed (%u!=%u)", | |
8a266467 TT |
4270 | g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)), |
4271 | le16_to_cpu(gdp->bg_checksum)); | |
4272 | err = -EINVAL; | |
4273 | goto restore_opts; | |
4274 | } | |
4275 | } | |
4276 | ||
ead6596b ES |
4277 | /* |
4278 | * If we have an unprocessed orphan list hanging | |
4279 | * around from a previously readonly bdev mount, | |
4280 | * require a full umount/remount for now. | |
4281 | */ | |
4282 | if (es->s_last_orphan) { | |
b31e1552 | 4283 | ext4_msg(sb, KERN_WARNING, "Couldn't " |
ead6596b ES |
4284 | "remount RDWR because of unprocessed " |
4285 | "orphan inode list. Please " | |
b31e1552 | 4286 | "umount/remount instead"); |
ead6596b ES |
4287 | err = -EINVAL; |
4288 | goto restore_opts; | |
4289 | } | |
4290 | ||
ac27a0ec DK |
4291 | /* |
4292 | * Mounting a RDONLY partition read-write, so reread | |
4293 | * and store the current valid flag. (It may have | |
4294 | * been changed by e2fsck since we originally mounted | |
4295 | * the partition.) | |
4296 | */ | |
0390131b FM |
4297 | if (sbi->s_journal) |
4298 | ext4_clear_journal_err(sb, es); | |
ac27a0ec | 4299 | sbi->s_mount_state = le16_to_cpu(es->s_state); |
617ba13b | 4300 | if ((err = ext4_group_extend(sb, es, n_blocks_count))) |
ac27a0ec | 4301 | goto restore_opts; |
2b2d6d01 | 4302 | if (!ext4_setup_super(sb, es, 0)) |
ac27a0ec | 4303 | sb->s_flags &= ~MS_RDONLY; |
c79d967d | 4304 | enable_quota = 1; |
ac27a0ec DK |
4305 | } |
4306 | } | |
bfff6873 LC |
4307 | |
4308 | /* | |
4309 | * Reinitialize lazy itable initialization thread based on | |
4310 | * current settings | |
4311 | */ | |
4312 | if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE)) | |
4313 | ext4_unregister_li_request(sb); | |
4314 | else { | |
4315 | ext4_group_t first_not_zeroed; | |
4316 | first_not_zeroed = ext4_has_uninit_itable(sb); | |
4317 | ext4_register_li_request(sb, first_not_zeroed); | |
4318 | } | |
4319 | ||
6fd058f7 | 4320 | ext4_setup_system_zone(sb); |
0390131b | 4321 | if (sbi->s_journal == NULL) |
e2d67052 | 4322 | ext4_commit_super(sb, 1); |
0390131b | 4323 | |
ac27a0ec DK |
4324 | #ifdef CONFIG_QUOTA |
4325 | /* Release old quota file names */ | |
4326 | for (i = 0; i < MAXQUOTAS; i++) | |
4327 | if (old_opts.s_qf_names[i] && | |
4328 | old_opts.s_qf_names[i] != sbi->s_qf_names[i]) | |
4329 | kfree(old_opts.s_qf_names[i]); | |
4330 | #endif | |
bbd6851a | 4331 | unlock_super(sb); |
c79d967d | 4332 | if (enable_quota) |
0f0dd62f | 4333 | dquot_resume(sb, -1); |
d4c402d9 CW |
4334 | |
4335 | ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); | |
4336 | kfree(orig_data); | |
ac27a0ec | 4337 | return 0; |
0b8e58a1 | 4338 | |
ac27a0ec DK |
4339 | restore_opts: |
4340 | sb->s_flags = old_sb_flags; | |
4341 | sbi->s_mount_opt = old_opts.s_mount_opt; | |
4342 | sbi->s_resuid = old_opts.s_resuid; | |
4343 | sbi->s_resgid = old_opts.s_resgid; | |
4344 | sbi->s_commit_interval = old_opts.s_commit_interval; | |
30773840 TT |
4345 | sbi->s_min_batch_time = old_opts.s_min_batch_time; |
4346 | sbi->s_max_batch_time = old_opts.s_max_batch_time; | |
ac27a0ec DK |
4347 | #ifdef CONFIG_QUOTA |
4348 | sbi->s_jquota_fmt = old_opts.s_jquota_fmt; | |
4349 | for (i = 0; i < MAXQUOTAS; i++) { | |
4350 | if (sbi->s_qf_names[i] && | |
4351 | old_opts.s_qf_names[i] != sbi->s_qf_names[i]) | |
4352 | kfree(sbi->s_qf_names[i]); | |
4353 | sbi->s_qf_names[i] = old_opts.s_qf_names[i]; | |
4354 | } | |
4355 | #endif | |
bbd6851a | 4356 | unlock_super(sb); |
d4c402d9 | 4357 | kfree(orig_data); |
ac27a0ec DK |
4358 | return err; |
4359 | } | |
4360 | ||
2b2d6d01 | 4361 | static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) |
ac27a0ec DK |
4362 | { |
4363 | struct super_block *sb = dentry->d_sb; | |
617ba13b MC |
4364 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
4365 | struct ext4_super_block *es = sbi->s_es; | |
960cc398 | 4366 | u64 fsid; |
ac27a0ec | 4367 | |
5e70030d BP |
4368 | if (test_opt(sb, MINIX_DF)) { |
4369 | sbi->s_overhead_last = 0; | |
6bc9feff | 4370 | } else if (sbi->s_blocks_last != ext4_blocks_count(es)) { |
8df9675f | 4371 | ext4_group_t i, ngroups = ext4_get_groups_count(sb); |
5e70030d | 4372 | ext4_fsblk_t overhead = 0; |
ac27a0ec DK |
4373 | |
4374 | /* | |
5e70030d BP |
4375 | * Compute the overhead (FS structures). This is constant |
4376 | * for a given filesystem unless the number of block groups | |
4377 | * changes so we cache the previous value until it does. | |
ac27a0ec DK |
4378 | */ |
4379 | ||
4380 | /* | |
4381 | * All of the blocks before first_data_block are | |
4382 | * overhead | |
4383 | */ | |
4384 | overhead = le32_to_cpu(es->s_first_data_block); | |
4385 | ||
4386 | /* | |
4387 | * Add the overhead attributed to the superblock and | |
4388 | * block group descriptors. If the sparse superblocks | |
4389 | * feature is turned on, then not all groups have this. | |
4390 | */ | |
4391 | for (i = 0; i < ngroups; i++) { | |
617ba13b MC |
4392 | overhead += ext4_bg_has_super(sb, i) + |
4393 | ext4_bg_num_gdb(sb, i); | |
ac27a0ec DK |
4394 | cond_resched(); |
4395 | } | |
4396 | ||
4397 | /* | |
4398 | * Every block group has an inode bitmap, a block | |
4399 | * bitmap, and an inode table. | |
4400 | */ | |
5e70030d BP |
4401 | overhead += ngroups * (2 + sbi->s_itb_per_group); |
4402 | sbi->s_overhead_last = overhead; | |
4403 | smp_wmb(); | |
6bc9feff | 4404 | sbi->s_blocks_last = ext4_blocks_count(es); |
ac27a0ec DK |
4405 | } |
4406 | ||
617ba13b | 4407 | buf->f_type = EXT4_SUPER_MAGIC; |
ac27a0ec | 4408 | buf->f_bsize = sb->s_blocksize; |
5e70030d | 4409 | buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last; |
6bc6e63f AK |
4410 | buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) - |
4411 | percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter); | |
bd81d8ee LV |
4412 | buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es); |
4413 | if (buf->f_bfree < ext4_r_blocks_count(es)) | |
ac27a0ec DK |
4414 | buf->f_bavail = 0; |
4415 | buf->f_files = le32_to_cpu(es->s_inodes_count); | |
52d9f3b4 | 4416 | buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); |
617ba13b | 4417 | buf->f_namelen = EXT4_NAME_LEN; |
960cc398 PE |
4418 | fsid = le64_to_cpup((void *)es->s_uuid) ^ |
4419 | le64_to_cpup((void *)es->s_uuid + sizeof(u64)); | |
4420 | buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; | |
4421 | buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; | |
0b8e58a1 | 4422 | |
ac27a0ec DK |
4423 | return 0; |
4424 | } | |
4425 | ||
0b8e58a1 AD |
4426 | /* Helper function for writing quotas on sync - we need to start transaction |
4427 | * before quota file is locked for write. Otherwise the are possible deadlocks: | |
ac27a0ec | 4428 | * Process 1 Process 2 |
617ba13b | 4429 | * ext4_create() quota_sync() |
a269eb18 | 4430 | * jbd2_journal_start() write_dquot() |
871a2931 | 4431 | * dquot_initialize() down(dqio_mutex) |
dab291af | 4432 | * down(dqio_mutex) jbd2_journal_start() |
ac27a0ec DK |
4433 | * |
4434 | */ | |
4435 | ||
4436 | #ifdef CONFIG_QUOTA | |
4437 | ||
4438 | static inline struct inode *dquot_to_inode(struct dquot *dquot) | |
4439 | { | |
4440 | return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type]; | |
4441 | } | |
4442 | ||
617ba13b | 4443 | static int ext4_write_dquot(struct dquot *dquot) |
ac27a0ec DK |
4444 | { |
4445 | int ret, err; | |
4446 | handle_t *handle; | |
4447 | struct inode *inode; | |
4448 | ||
4449 | inode = dquot_to_inode(dquot); | |
617ba13b | 4450 | handle = ext4_journal_start(inode, |
0b8e58a1 | 4451 | EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); |
ac27a0ec DK |
4452 | if (IS_ERR(handle)) |
4453 | return PTR_ERR(handle); | |
4454 | ret = dquot_commit(dquot); | |
617ba13b | 4455 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
4456 | if (!ret) |
4457 | ret = err; | |
4458 | return ret; | |
4459 | } | |
4460 | ||
617ba13b | 4461 | static int ext4_acquire_dquot(struct dquot *dquot) |
ac27a0ec DK |
4462 | { |
4463 | int ret, err; | |
4464 | handle_t *handle; | |
4465 | ||
617ba13b | 4466 | handle = ext4_journal_start(dquot_to_inode(dquot), |
0b8e58a1 | 4467 | EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); |
ac27a0ec DK |
4468 | if (IS_ERR(handle)) |
4469 | return PTR_ERR(handle); | |
4470 | ret = dquot_acquire(dquot); | |
617ba13b | 4471 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
4472 | if (!ret) |
4473 | ret = err; | |
4474 | return ret; | |
4475 | } | |
4476 | ||
617ba13b | 4477 | static int ext4_release_dquot(struct dquot *dquot) |
ac27a0ec DK |
4478 | { |
4479 | int ret, err; | |
4480 | handle_t *handle; | |
4481 | ||
617ba13b | 4482 | handle = ext4_journal_start(dquot_to_inode(dquot), |
0b8e58a1 | 4483 | EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); |
9c3013e9 JK |
4484 | if (IS_ERR(handle)) { |
4485 | /* Release dquot anyway to avoid endless cycle in dqput() */ | |
4486 | dquot_release(dquot); | |
ac27a0ec | 4487 | return PTR_ERR(handle); |
9c3013e9 | 4488 | } |
ac27a0ec | 4489 | ret = dquot_release(dquot); |
617ba13b | 4490 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
4491 | if (!ret) |
4492 | ret = err; | |
4493 | return ret; | |
4494 | } | |
4495 | ||
617ba13b | 4496 | static int ext4_mark_dquot_dirty(struct dquot *dquot) |
ac27a0ec | 4497 | { |
2c8be6b2 | 4498 | /* Are we journaling quotas? */ |
617ba13b MC |
4499 | if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] || |
4500 | EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) { | |
ac27a0ec | 4501 | dquot_mark_dquot_dirty(dquot); |
617ba13b | 4502 | return ext4_write_dquot(dquot); |
ac27a0ec DK |
4503 | } else { |
4504 | return dquot_mark_dquot_dirty(dquot); | |
4505 | } | |
4506 | } | |
4507 | ||
617ba13b | 4508 | static int ext4_write_info(struct super_block *sb, int type) |
ac27a0ec DK |
4509 | { |
4510 | int ret, err; | |
4511 | handle_t *handle; | |
4512 | ||
4513 | /* Data block + inode block */ | |
617ba13b | 4514 | handle = ext4_journal_start(sb->s_root->d_inode, 2); |
ac27a0ec DK |
4515 | if (IS_ERR(handle)) |
4516 | return PTR_ERR(handle); | |
4517 | ret = dquot_commit_info(sb, type); | |
617ba13b | 4518 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
4519 | if (!ret) |
4520 | ret = err; | |
4521 | return ret; | |
4522 | } | |
4523 | ||
4524 | /* | |
4525 | * Turn on quotas during mount time - we need to find | |
4526 | * the quota file and such... | |
4527 | */ | |
617ba13b | 4528 | static int ext4_quota_on_mount(struct super_block *sb, int type) |
ac27a0ec | 4529 | { |
287a8095 CH |
4530 | return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type], |
4531 | EXT4_SB(sb)->s_jquota_fmt, type); | |
ac27a0ec DK |
4532 | } |
4533 | ||
4534 | /* | |
4535 | * Standard function to be called on quota_on | |
4536 | */ | |
617ba13b | 4537 | static int ext4_quota_on(struct super_block *sb, int type, int format_id, |
307ae18a | 4538 | char *name) |
ac27a0ec DK |
4539 | { |
4540 | int err; | |
8264613d | 4541 | struct path path; |
ac27a0ec DK |
4542 | |
4543 | if (!test_opt(sb, QUOTA)) | |
4544 | return -EINVAL; | |
0623543b | 4545 | |
8264613d | 4546 | err = kern_path(name, LOOKUP_FOLLOW, &path); |
ac27a0ec DK |
4547 | if (err) |
4548 | return err; | |
0623543b | 4549 | |
ac27a0ec | 4550 | /* Quotafile not on the same filesystem? */ |
8264613d AV |
4551 | if (path.mnt->mnt_sb != sb) { |
4552 | path_put(&path); | |
ac27a0ec DK |
4553 | return -EXDEV; |
4554 | } | |
0623543b JK |
4555 | /* Journaling quota? */ |
4556 | if (EXT4_SB(sb)->s_qf_names[type]) { | |
2b2d6d01 | 4557 | /* Quotafile not in fs root? */ |
8264613d | 4558 | if (path.dentry->d_parent != sb->s_root) |
b31e1552 ES |
4559 | ext4_msg(sb, KERN_WARNING, |
4560 | "Quota file not on filesystem root. " | |
4561 | "Journaled quota will not work"); | |
2b2d6d01 | 4562 | } |
0623543b JK |
4563 | |
4564 | /* | |
4565 | * When we journal data on quota file, we have to flush journal to see | |
4566 | * all updates to the file when we bypass pagecache... | |
4567 | */ | |
0390131b FM |
4568 | if (EXT4_SB(sb)->s_journal && |
4569 | ext4_should_journal_data(path.dentry->d_inode)) { | |
0623543b JK |
4570 | /* |
4571 | * We don't need to lock updates but journal_flush() could | |
4572 | * otherwise be livelocked... | |
4573 | */ | |
4574 | jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); | |
7ffe1ea8 | 4575 | err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); |
0623543b | 4576 | jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); |
7ffe1ea8 | 4577 | if (err) { |
8264613d | 4578 | path_put(&path); |
7ffe1ea8 HK |
4579 | return err; |
4580 | } | |
0623543b JK |
4581 | } |
4582 | ||
287a8095 | 4583 | err = dquot_quota_on_path(sb, type, format_id, &path); |
8264613d | 4584 | path_put(&path); |
77e69dac | 4585 | return err; |
ac27a0ec DK |
4586 | } |
4587 | ||
ca0e05e4 DM |
4588 | static int ext4_quota_off(struct super_block *sb, int type) |
4589 | { | |
87009d86 DM |
4590 | /* Force all delayed allocation blocks to be allocated. |
4591 | * Caller already holds s_umount sem */ | |
4592 | if (test_opt(sb, DELALLOC)) | |
ca0e05e4 | 4593 | sync_filesystem(sb); |
ca0e05e4 DM |
4594 | |
4595 | return dquot_quota_off(sb, type); | |
4596 | } | |
4597 | ||
ac27a0ec DK |
4598 | /* Read data from quotafile - avoid pagecache and such because we cannot afford |
4599 | * acquiring the locks... As quota files are never truncated and quota code | |
4600 | * itself serializes the operations (and noone else should touch the files) | |
4601 | * we don't have to be afraid of races */ | |
617ba13b | 4602 | static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, |
ac27a0ec DK |
4603 | size_t len, loff_t off) |
4604 | { | |
4605 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
725d26d3 | 4606 | ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); |
ac27a0ec DK |
4607 | int err = 0; |
4608 | int offset = off & (sb->s_blocksize - 1); | |
4609 | int tocopy; | |
4610 | size_t toread; | |
4611 | struct buffer_head *bh; | |
4612 | loff_t i_size = i_size_read(inode); | |
4613 | ||
4614 | if (off > i_size) | |
4615 | return 0; | |
4616 | if (off+len > i_size) | |
4617 | len = i_size-off; | |
4618 | toread = len; | |
4619 | while (toread > 0) { | |
4620 | tocopy = sb->s_blocksize - offset < toread ? | |
4621 | sb->s_blocksize - offset : toread; | |
617ba13b | 4622 | bh = ext4_bread(NULL, inode, blk, 0, &err); |
ac27a0ec DK |
4623 | if (err) |
4624 | return err; | |
4625 | if (!bh) /* A hole? */ | |
4626 | memset(data, 0, tocopy); | |
4627 | else | |
4628 | memcpy(data, bh->b_data+offset, tocopy); | |
4629 | brelse(bh); | |
4630 | offset = 0; | |
4631 | toread -= tocopy; | |
4632 | data += tocopy; | |
4633 | blk++; | |
4634 | } | |
4635 | return len; | |
4636 | } | |
4637 | ||
4638 | /* Write to quotafile (we know the transaction is already started and has | |
4639 | * enough credits) */ | |
617ba13b | 4640 | static ssize_t ext4_quota_write(struct super_block *sb, int type, |
ac27a0ec DK |
4641 | const char *data, size_t len, loff_t off) |
4642 | { | |
4643 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
725d26d3 | 4644 | ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); |
ac27a0ec DK |
4645 | int err = 0; |
4646 | int offset = off & (sb->s_blocksize - 1); | |
ac27a0ec DK |
4647 | struct buffer_head *bh; |
4648 | handle_t *handle = journal_current_handle(); | |
4649 | ||
0390131b | 4650 | if (EXT4_SB(sb)->s_journal && !handle) { |
b31e1552 ES |
4651 | ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" |
4652 | " cancelled because transaction is not started", | |
9c3013e9 JK |
4653 | (unsigned long long)off, (unsigned long long)len); |
4654 | return -EIO; | |
4655 | } | |
67eeb568 DM |
4656 | /* |
4657 | * Since we account only one data block in transaction credits, | |
4658 | * then it is impossible to cross a block boundary. | |
4659 | */ | |
4660 | if (sb->s_blocksize - offset < len) { | |
4661 | ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" | |
4662 | " cancelled because not block aligned", | |
4663 | (unsigned long long)off, (unsigned long long)len); | |
4664 | return -EIO; | |
4665 | } | |
4666 | ||
ac27a0ec | 4667 | mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA); |
67eeb568 DM |
4668 | bh = ext4_bread(handle, inode, blk, 1, &err); |
4669 | if (!bh) | |
4670 | goto out; | |
62d2b5f2 JK |
4671 | err = ext4_journal_get_write_access(handle, bh); |
4672 | if (err) { | |
4673 | brelse(bh); | |
4674 | goto out; | |
ac27a0ec | 4675 | } |
67eeb568 DM |
4676 | lock_buffer(bh); |
4677 | memcpy(bh->b_data+offset, data, len); | |
4678 | flush_dcache_page(bh->b_page); | |
4679 | unlock_buffer(bh); | |
62d2b5f2 | 4680 | err = ext4_handle_dirty_metadata(handle, NULL, bh); |
67eeb568 | 4681 | brelse(bh); |
ac27a0ec | 4682 | out: |
67eeb568 | 4683 | if (err) { |
4d04e4fb | 4684 | mutex_unlock(&inode->i_mutex); |
ac27a0ec | 4685 | return err; |
4d04e4fb | 4686 | } |
67eeb568 DM |
4687 | if (inode->i_size < off + len) { |
4688 | i_size_write(inode, off + len); | |
617ba13b | 4689 | EXT4_I(inode)->i_disksize = inode->i_size; |
ac27a0ec | 4690 | } |
ac27a0ec | 4691 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; |
617ba13b | 4692 | ext4_mark_inode_dirty(handle, inode); |
ac27a0ec | 4693 | mutex_unlock(&inode->i_mutex); |
67eeb568 | 4694 | return len; |
ac27a0ec DK |
4695 | } |
4696 | ||
4697 | #endif | |
4698 | ||
152a0836 AV |
4699 | static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, |
4700 | const char *dev_name, void *data) | |
ac27a0ec | 4701 | { |
152a0836 | 4702 | return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); |
ac27a0ec DK |
4703 | } |
4704 | ||
37f328eb | 4705 | #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) |
24b58424 TT |
4706 | static struct file_system_type ext2_fs_type = { |
4707 | .owner = THIS_MODULE, | |
4708 | .name = "ext2", | |
152a0836 | 4709 | .mount = ext4_mount, |
24b58424 TT |
4710 | .kill_sb = kill_block_super, |
4711 | .fs_flags = FS_REQUIRES_DEV, | |
4712 | }; | |
4713 | ||
4714 | static inline void register_as_ext2(void) | |
4715 | { | |
4716 | int err = register_filesystem(&ext2_fs_type); | |
4717 | if (err) | |
4718 | printk(KERN_WARNING | |
4719 | "EXT4-fs: Unable to register as ext2 (%d)\n", err); | |
4720 | } | |
4721 | ||
4722 | static inline void unregister_as_ext2(void) | |
4723 | { | |
4724 | unregister_filesystem(&ext2_fs_type); | |
4725 | } | |
51b7e3c9 | 4726 | MODULE_ALIAS("ext2"); |
24b58424 TT |
4727 | #else |
4728 | static inline void register_as_ext2(void) { } | |
4729 | static inline void unregister_as_ext2(void) { } | |
4730 | #endif | |
4731 | ||
37f328eb | 4732 | #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) |
24b58424 TT |
4733 | static inline void register_as_ext3(void) |
4734 | { | |
4735 | int err = register_filesystem(&ext3_fs_type); | |
4736 | if (err) | |
4737 | printk(KERN_WARNING | |
4738 | "EXT4-fs: Unable to register as ext3 (%d)\n", err); | |
4739 | } | |
4740 | ||
4741 | static inline void unregister_as_ext3(void) | |
4742 | { | |
4743 | unregister_filesystem(&ext3_fs_type); | |
4744 | } | |
51b7e3c9 | 4745 | MODULE_ALIAS("ext3"); |
24b58424 TT |
4746 | #else |
4747 | static inline void register_as_ext3(void) { } | |
4748 | static inline void unregister_as_ext3(void) { } | |
4749 | #endif | |
4750 | ||
03010a33 TT |
4751 | static struct file_system_type ext4_fs_type = { |
4752 | .owner = THIS_MODULE, | |
4753 | .name = "ext4", | |
152a0836 | 4754 | .mount = ext4_mount, |
03010a33 TT |
4755 | .kill_sb = kill_block_super, |
4756 | .fs_flags = FS_REQUIRES_DEV, | |
4757 | }; | |
4758 | ||
857ac889 LC |
4759 | int __init ext4_init_feat_adverts(void) |
4760 | { | |
4761 | struct ext4_features *ef; | |
4762 | int ret = -ENOMEM; | |
4763 | ||
4764 | ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL); | |
4765 | if (!ef) | |
4766 | goto out; | |
4767 | ||
4768 | ef->f_kobj.kset = ext4_kset; | |
4769 | init_completion(&ef->f_kobj_unregister); | |
4770 | ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL, | |
4771 | "features"); | |
4772 | if (ret) { | |
4773 | kfree(ef); | |
4774 | goto out; | |
4775 | } | |
4776 | ||
4777 | ext4_feat = ef; | |
4778 | ret = 0; | |
4779 | out: | |
4780 | return ret; | |
4781 | } | |
4782 | ||
5dabfc78 | 4783 | static int __init ext4_init_fs(void) |
ac27a0ec | 4784 | { |
c9de560d AT |
4785 | int err; |
4786 | ||
12e9b892 | 4787 | ext4_check_flag_values(); |
5dabfc78 | 4788 | err = ext4_init_pageio(); |
6fd058f7 TT |
4789 | if (err) |
4790 | return err; | |
5dabfc78 | 4791 | err = ext4_init_system_zone(); |
bd2d0210 TT |
4792 | if (err) |
4793 | goto out5; | |
3197ebdb TT |
4794 | ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj); |
4795 | if (!ext4_kset) | |
6fd058f7 | 4796 | goto out4; |
9f6200bb | 4797 | ext4_proc_root = proc_mkdir("fs/ext4", NULL); |
857ac889 LC |
4798 | |
4799 | err = ext4_init_feat_adverts(); | |
4800 | ||
5dabfc78 | 4801 | err = ext4_init_mballoc(); |
ac27a0ec | 4802 | if (err) |
6fd058f7 | 4803 | goto out3; |
c9de560d | 4804 | |
5dabfc78 | 4805 | err = ext4_init_xattr(); |
c9de560d AT |
4806 | if (err) |
4807 | goto out2; | |
ac27a0ec DK |
4808 | err = init_inodecache(); |
4809 | if (err) | |
4810 | goto out1; | |
24b58424 TT |
4811 | register_as_ext2(); |
4812 | register_as_ext3(); | |
03010a33 | 4813 | err = register_filesystem(&ext4_fs_type); |
ac27a0ec DK |
4814 | if (err) |
4815 | goto out; | |
bfff6873 LC |
4816 | |
4817 | ext4_li_info = NULL; | |
4818 | mutex_init(&ext4_li_mtx); | |
ac27a0ec DK |
4819 | return 0; |
4820 | out: | |
24b58424 TT |
4821 | unregister_as_ext2(); |
4822 | unregister_as_ext3(); | |
ac27a0ec DK |
4823 | destroy_inodecache(); |
4824 | out1: | |
5dabfc78 | 4825 | ext4_exit_xattr(); |
c9de560d | 4826 | out2: |
5dabfc78 | 4827 | ext4_exit_mballoc(); |
6fd058f7 | 4828 | out3: |
857ac889 | 4829 | kfree(ext4_feat); |
6fd058f7 TT |
4830 | remove_proc_entry("fs/ext4", NULL); |
4831 | kset_unregister(ext4_kset); | |
4832 | out4: | |
5dabfc78 | 4833 | ext4_exit_system_zone(); |
bd2d0210 | 4834 | out5: |
5dabfc78 | 4835 | ext4_exit_pageio(); |
ac27a0ec DK |
4836 | return err; |
4837 | } | |
4838 | ||
5dabfc78 | 4839 | static void __exit ext4_exit_fs(void) |
ac27a0ec | 4840 | { |
bfff6873 | 4841 | ext4_destroy_lazyinit_thread(); |
24b58424 TT |
4842 | unregister_as_ext2(); |
4843 | unregister_as_ext3(); | |
03010a33 | 4844 | unregister_filesystem(&ext4_fs_type); |
ac27a0ec | 4845 | destroy_inodecache(); |
5dabfc78 TT |
4846 | ext4_exit_xattr(); |
4847 | ext4_exit_mballoc(); | |
9f6200bb | 4848 | remove_proc_entry("fs/ext4", NULL); |
3197ebdb | 4849 | kset_unregister(ext4_kset); |
5dabfc78 TT |
4850 | ext4_exit_system_zone(); |
4851 | ext4_exit_pageio(); | |
ac27a0ec DK |
4852 | } |
4853 | ||
4854 | MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); | |
83982b6f | 4855 | MODULE_DESCRIPTION("Fourth Extended Filesystem"); |
ac27a0ec | 4856 | MODULE_LICENSE("GPL"); |
5dabfc78 TT |
4857 | module_init(ext4_init_fs) |
4858 | module_exit(ext4_exit_fs) |