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> | |
29 | #include <linux/smp_lock.h> | |
30 | #include <linux/buffer_head.h> | |
a5694255 | 31 | #include <linux/exportfs.h> |
ac27a0ec DK |
32 | #include <linux/vfs.h> |
33 | #include <linux/random.h> | |
34 | #include <linux/mount.h> | |
35 | #include <linux/namei.h> | |
36 | #include <linux/quotaops.h> | |
37 | #include <linux/seq_file.h> | |
9f6200bb | 38 | #include <linux/proc_fs.h> |
3197ebdb | 39 | #include <linux/ctype.h> |
1330593e | 40 | #include <linux/log2.h> |
717d50e4 | 41 | #include <linux/crc16.h> |
ac27a0ec DK |
42 | #include <asm/uaccess.h> |
43 | ||
3dcf5451 CH |
44 | #include "ext4.h" |
45 | #include "ext4_jbd2.h" | |
ac27a0ec DK |
46 | #include "xattr.h" |
47 | #include "acl.h" | |
3661d286 | 48 | #include "mballoc.h" |
ac27a0ec | 49 | |
9bffad1e TT |
50 | #define CREATE_TRACE_POINTS |
51 | #include <trace/events/ext4.h> | |
52 | ||
f4033903 CW |
53 | static int default_mb_history_length = 1000; |
54 | ||
55 | module_param_named(default_mb_history_length, default_mb_history_length, | |
56 | int, 0644); | |
57 | MODULE_PARM_DESC(default_mb_history_length, | |
58 | "Default number of entries saved for mb_history"); | |
59 | ||
9f6200bb | 60 | struct proc_dir_entry *ext4_proc_root; |
3197ebdb | 61 | static struct kset *ext4_kset; |
9f6200bb | 62 | |
617ba13b | 63 | static int ext4_load_journal(struct super_block *, struct ext4_super_block *, |
ac27a0ec | 64 | unsigned long journal_devnum); |
e2d67052 | 65 | static int ext4_commit_super(struct super_block *sb, int sync); |
2b2d6d01 TT |
66 | static void ext4_mark_recovery_complete(struct super_block *sb, |
67 | struct ext4_super_block *es); | |
68 | static void ext4_clear_journal_err(struct super_block *sb, | |
69 | struct ext4_super_block *es); | |
617ba13b | 70 | static int ext4_sync_fs(struct super_block *sb, int wait); |
2b2d6d01 | 71 | static const char *ext4_decode_error(struct super_block *sb, int errno, |
ac27a0ec | 72 | char nbuf[16]); |
2b2d6d01 TT |
73 | static int ext4_remount(struct super_block *sb, int *flags, char *data); |
74 | static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); | |
c4be0c1d | 75 | static int ext4_unfreeze(struct super_block *sb); |
2b2d6d01 | 76 | static void ext4_write_super(struct super_block *sb); |
c4be0c1d | 77 | static int ext4_freeze(struct super_block *sb); |
ac27a0ec | 78 | |
bd81d8ee | 79 | |
8fadc143 AR |
80 | ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, |
81 | struct ext4_group_desc *bg) | |
bd81d8ee | 82 | { |
3a14589c | 83 | return le32_to_cpu(bg->bg_block_bitmap_lo) | |
8fadc143 | 84 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 85 | (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); |
bd81d8ee LV |
86 | } |
87 | ||
8fadc143 AR |
88 | ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, |
89 | struct ext4_group_desc *bg) | |
bd81d8ee | 90 | { |
5272f837 | 91 | return le32_to_cpu(bg->bg_inode_bitmap_lo) | |
8fadc143 | 92 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 93 | (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); |
bd81d8ee LV |
94 | } |
95 | ||
8fadc143 AR |
96 | ext4_fsblk_t ext4_inode_table(struct super_block *sb, |
97 | struct ext4_group_desc *bg) | |
bd81d8ee | 98 | { |
5272f837 | 99 | return le32_to_cpu(bg->bg_inode_table_lo) | |
8fadc143 | 100 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 101 | (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); |
bd81d8ee LV |
102 | } |
103 | ||
560671a0 AK |
104 | __u32 ext4_free_blks_count(struct super_block *sb, |
105 | struct ext4_group_desc *bg) | |
106 | { | |
107 | return le16_to_cpu(bg->bg_free_blocks_count_lo) | | |
108 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 109 | (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); |
560671a0 AK |
110 | } |
111 | ||
112 | __u32 ext4_free_inodes_count(struct super_block *sb, | |
113 | struct ext4_group_desc *bg) | |
114 | { | |
115 | return le16_to_cpu(bg->bg_free_inodes_count_lo) | | |
116 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 117 | (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); |
560671a0 AK |
118 | } |
119 | ||
120 | __u32 ext4_used_dirs_count(struct super_block *sb, | |
121 | struct ext4_group_desc *bg) | |
122 | { | |
123 | return le16_to_cpu(bg->bg_used_dirs_count_lo) | | |
124 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 125 | (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); |
560671a0 AK |
126 | } |
127 | ||
128 | __u32 ext4_itable_unused_count(struct super_block *sb, | |
129 | struct ext4_group_desc *bg) | |
130 | { | |
131 | return le16_to_cpu(bg->bg_itable_unused_lo) | | |
132 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 133 | (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); |
560671a0 AK |
134 | } |
135 | ||
8fadc143 AR |
136 | void ext4_block_bitmap_set(struct super_block *sb, |
137 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 138 | { |
3a14589c | 139 | bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
140 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
141 | bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
142 | } |
143 | ||
8fadc143 AR |
144 | void ext4_inode_bitmap_set(struct super_block *sb, |
145 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 146 | { |
5272f837 | 147 | bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
148 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
149 | bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
150 | } |
151 | ||
8fadc143 AR |
152 | void ext4_inode_table_set(struct super_block *sb, |
153 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 154 | { |
5272f837 | 155 | bg->bg_inode_table_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
156 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
157 | bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
158 | } |
159 | ||
560671a0 AK |
160 | void ext4_free_blks_set(struct super_block *sb, |
161 | struct ext4_group_desc *bg, __u32 count) | |
162 | { | |
163 | bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); | |
164 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
165 | bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); | |
166 | } | |
167 | ||
168 | void ext4_free_inodes_set(struct super_block *sb, | |
169 | struct ext4_group_desc *bg, __u32 count) | |
170 | { | |
171 | bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); | |
172 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
173 | bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); | |
174 | } | |
175 | ||
176 | void ext4_used_dirs_set(struct super_block *sb, | |
177 | struct ext4_group_desc *bg, __u32 count) | |
178 | { | |
179 | bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); | |
180 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
181 | bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); | |
182 | } | |
183 | ||
184 | void ext4_itable_unused_set(struct super_block *sb, | |
185 | struct ext4_group_desc *bg, __u32 count) | |
186 | { | |
187 | bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); | |
188 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
189 | bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); | |
190 | } | |
191 | ||
ac27a0ec | 192 | /* |
dab291af | 193 | * Wrappers for jbd2_journal_start/end. |
ac27a0ec DK |
194 | * |
195 | * The only special thing we need to do here is to make sure that all | |
196 | * journal_end calls result in the superblock being marked dirty, so | |
197 | * that sync() will call the filesystem's write_super callback if | |
198 | * appropriate. | |
199 | */ | |
617ba13b | 200 | handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks) |
ac27a0ec DK |
201 | { |
202 | journal_t *journal; | |
203 | ||
204 | if (sb->s_flags & MS_RDONLY) | |
205 | return ERR_PTR(-EROFS); | |
206 | ||
207 | /* Special case here: if the journal has aborted behind our | |
208 | * backs (eg. EIO in the commit thread), then we still need to | |
209 | * take the FS itself readonly cleanly. */ | |
617ba13b | 210 | journal = EXT4_SB(sb)->s_journal; |
0390131b FM |
211 | if (journal) { |
212 | if (is_journal_aborted(journal)) { | |
0b8e58a1 | 213 | ext4_abort(sb, __func__, "Detected aborted journal"); |
0390131b FM |
214 | return ERR_PTR(-EROFS); |
215 | } | |
216 | return jbd2_journal_start(journal, nblocks); | |
ac27a0ec | 217 | } |
0390131b FM |
218 | /* |
219 | * We're not journaling, return the appropriate indication. | |
220 | */ | |
221 | current->journal_info = EXT4_NOJOURNAL_HANDLE; | |
222 | return current->journal_info; | |
ac27a0ec DK |
223 | } |
224 | ||
225 | /* | |
226 | * The only special thing we need to do here is to make sure that all | |
dab291af | 227 | * jbd2_journal_stop calls result in the superblock being marked dirty, so |
ac27a0ec DK |
228 | * that sync() will call the filesystem's write_super callback if |
229 | * appropriate. | |
230 | */ | |
617ba13b | 231 | int __ext4_journal_stop(const char *where, handle_t *handle) |
ac27a0ec DK |
232 | { |
233 | struct super_block *sb; | |
234 | int err; | |
235 | int rc; | |
236 | ||
0390131b FM |
237 | if (!ext4_handle_valid(handle)) { |
238 | /* | |
239 | * Do this here since we don't call jbd2_journal_stop() in | |
240 | * no-journal mode. | |
241 | */ | |
242 | current->journal_info = NULL; | |
243 | return 0; | |
244 | } | |
ac27a0ec DK |
245 | sb = handle->h_transaction->t_journal->j_private; |
246 | err = handle->h_err; | |
dab291af | 247 | rc = jbd2_journal_stop(handle); |
ac27a0ec DK |
248 | |
249 | if (!err) | |
250 | err = rc; | |
251 | if (err) | |
617ba13b | 252 | __ext4_std_error(sb, where, err); |
ac27a0ec DK |
253 | return err; |
254 | } | |
255 | ||
617ba13b | 256 | void ext4_journal_abort_handle(const char *caller, const char *err_fn, |
ac27a0ec DK |
257 | struct buffer_head *bh, handle_t *handle, int err) |
258 | { | |
259 | char nbuf[16]; | |
617ba13b | 260 | const char *errstr = ext4_decode_error(NULL, err, nbuf); |
ac27a0ec | 261 | |
0390131b FM |
262 | BUG_ON(!ext4_handle_valid(handle)); |
263 | ||
ac27a0ec DK |
264 | if (bh) |
265 | BUFFER_TRACE(bh, "abort"); | |
266 | ||
267 | if (!handle->h_err) | |
268 | handle->h_err = err; | |
269 | ||
270 | if (is_handle_aborted(handle)) | |
271 | return; | |
272 | ||
273 | printk(KERN_ERR "%s: aborting transaction: %s in %s\n", | |
274 | caller, errstr, err_fn); | |
275 | ||
dab291af | 276 | jbd2_journal_abort_handle(handle); |
ac27a0ec DK |
277 | } |
278 | ||
279 | /* Deal with the reporting of failure conditions on a filesystem such as | |
280 | * inconsistencies detected or read IO failures. | |
281 | * | |
282 | * On ext2, we can store the error state of the filesystem in the | |
617ba13b | 283 | * superblock. That is not possible on ext4, because we may have other |
ac27a0ec DK |
284 | * write ordering constraints on the superblock which prevent us from |
285 | * writing it out straight away; and given that the journal is about to | |
286 | * be aborted, we can't rely on the current, or future, transactions to | |
287 | * write out the superblock safely. | |
288 | * | |
dab291af | 289 | * We'll just use the jbd2_journal_abort() error code to record an error in |
ac27a0ec DK |
290 | * the journal instead. On recovery, the journal will compain about |
291 | * that error until we've noted it down and cleared it. | |
292 | */ | |
293 | ||
617ba13b | 294 | static void ext4_handle_error(struct super_block *sb) |
ac27a0ec | 295 | { |
617ba13b | 296 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
ac27a0ec | 297 | |
617ba13b MC |
298 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; |
299 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
ac27a0ec DK |
300 | |
301 | if (sb->s_flags & MS_RDONLY) | |
302 | return; | |
303 | ||
2b2d6d01 | 304 | if (!test_opt(sb, ERRORS_CONT)) { |
617ba13b | 305 | journal_t *journal = EXT4_SB(sb)->s_journal; |
ac27a0ec | 306 | |
4ab2f15b | 307 | EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; |
ac27a0ec | 308 | if (journal) |
dab291af | 309 | jbd2_journal_abort(journal, -EIO); |
ac27a0ec | 310 | } |
2b2d6d01 | 311 | if (test_opt(sb, ERRORS_RO)) { |
b31e1552 | 312 | ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); |
ac27a0ec DK |
313 | sb->s_flags |= MS_RDONLY; |
314 | } | |
e2d67052 | 315 | ext4_commit_super(sb, 1); |
ac27a0ec | 316 | if (test_opt(sb, ERRORS_PANIC)) |
617ba13b | 317 | panic("EXT4-fs (device %s): panic forced after error\n", |
ac27a0ec DK |
318 | sb->s_id); |
319 | } | |
320 | ||
2b2d6d01 TT |
321 | void ext4_error(struct super_block *sb, const char *function, |
322 | const char *fmt, ...) | |
ac27a0ec DK |
323 | { |
324 | va_list args; | |
325 | ||
326 | va_start(args, fmt); | |
2b2d6d01 | 327 | printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function); |
ac27a0ec DK |
328 | vprintk(fmt, args); |
329 | printk("\n"); | |
330 | va_end(args); | |
331 | ||
617ba13b | 332 | ext4_handle_error(sb); |
ac27a0ec DK |
333 | } |
334 | ||
2b2d6d01 | 335 | static const char *ext4_decode_error(struct super_block *sb, int errno, |
ac27a0ec DK |
336 | char nbuf[16]) |
337 | { | |
338 | char *errstr = NULL; | |
339 | ||
340 | switch (errno) { | |
341 | case -EIO: | |
342 | errstr = "IO failure"; | |
343 | break; | |
344 | case -ENOMEM: | |
345 | errstr = "Out of memory"; | |
346 | break; | |
347 | case -EROFS: | |
78f1ddbb TT |
348 | if (!sb || (EXT4_SB(sb)->s_journal && |
349 | EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) | |
ac27a0ec DK |
350 | errstr = "Journal has aborted"; |
351 | else | |
352 | errstr = "Readonly filesystem"; | |
353 | break; | |
354 | default: | |
355 | /* If the caller passed in an extra buffer for unknown | |
356 | * errors, textualise them now. Else we just return | |
357 | * NULL. */ | |
358 | if (nbuf) { | |
359 | /* Check for truncated error codes... */ | |
360 | if (snprintf(nbuf, 16, "error %d", -errno) >= 0) | |
361 | errstr = nbuf; | |
362 | } | |
363 | break; | |
364 | } | |
365 | ||
366 | return errstr; | |
367 | } | |
368 | ||
617ba13b | 369 | /* __ext4_std_error decodes expected errors from journaling functions |
ac27a0ec DK |
370 | * automatically and invokes the appropriate error response. */ |
371 | ||
2b2d6d01 | 372 | void __ext4_std_error(struct super_block *sb, const char *function, int errno) |
ac27a0ec DK |
373 | { |
374 | char nbuf[16]; | |
375 | const char *errstr; | |
376 | ||
377 | /* Special case: if the error is EROFS, and we're not already | |
378 | * inside a transaction, then there's really no point in logging | |
379 | * an error. */ | |
380 | if (errno == -EROFS && journal_current_handle() == NULL && | |
381 | (sb->s_flags & MS_RDONLY)) | |
382 | return; | |
383 | ||
617ba13b | 384 | errstr = ext4_decode_error(sb, errno, nbuf); |
2b2d6d01 TT |
385 | printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n", |
386 | sb->s_id, function, errstr); | |
ac27a0ec | 387 | |
617ba13b | 388 | ext4_handle_error(sb); |
ac27a0ec DK |
389 | } |
390 | ||
391 | /* | |
617ba13b | 392 | * ext4_abort is a much stronger failure handler than ext4_error. The |
ac27a0ec DK |
393 | * abort function may be used to deal with unrecoverable failures such |
394 | * as journal IO errors or ENOMEM at a critical moment in log management. | |
395 | * | |
396 | * We unconditionally force the filesystem into an ABORT|READONLY state, | |
397 | * unless the error response on the fs has been set to panic in which | |
398 | * case we take the easy way out and panic immediately. | |
399 | */ | |
400 | ||
2b2d6d01 TT |
401 | void ext4_abort(struct super_block *sb, const char *function, |
402 | const char *fmt, ...) | |
ac27a0ec DK |
403 | { |
404 | va_list args; | |
405 | ||
ac27a0ec | 406 | va_start(args, fmt); |
2b2d6d01 | 407 | printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function); |
ac27a0ec DK |
408 | vprintk(fmt, args); |
409 | printk("\n"); | |
410 | va_end(args); | |
411 | ||
412 | if (test_opt(sb, ERRORS_PANIC)) | |
617ba13b | 413 | panic("EXT4-fs panic from previous error\n"); |
ac27a0ec DK |
414 | |
415 | if (sb->s_flags & MS_RDONLY) | |
416 | return; | |
417 | ||
b31e1552 | 418 | ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); |
617ba13b | 419 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; |
ac27a0ec | 420 | sb->s_flags |= MS_RDONLY; |
4ab2f15b | 421 | EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; |
ef2cabf7 HK |
422 | if (EXT4_SB(sb)->s_journal) |
423 | jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); | |
ac27a0ec DK |
424 | } |
425 | ||
b31e1552 ES |
426 | void ext4_msg (struct super_block * sb, const char *prefix, |
427 | const char *fmt, ...) | |
428 | { | |
429 | va_list args; | |
430 | ||
431 | va_start(args, fmt); | |
432 | printk("%sEXT4-fs (%s): ", prefix, sb->s_id); | |
433 | vprintk(fmt, args); | |
434 | printk("\n"); | |
435 | va_end(args); | |
436 | } | |
437 | ||
2b2d6d01 TT |
438 | void ext4_warning(struct super_block *sb, const char *function, |
439 | const char *fmt, ...) | |
ac27a0ec DK |
440 | { |
441 | va_list args; | |
442 | ||
443 | va_start(args, fmt); | |
617ba13b | 444 | printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ", |
ac27a0ec DK |
445 | sb->s_id, function); |
446 | vprintk(fmt, args); | |
447 | printk("\n"); | |
448 | va_end(args); | |
449 | } | |
450 | ||
5d1b1b3f | 451 | void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp, |
0b8e58a1 | 452 | const char *function, const char *fmt, ...) |
5d1b1b3f AK |
453 | __releases(bitlock) |
454 | __acquires(bitlock) | |
455 | { | |
456 | va_list args; | |
457 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; | |
458 | ||
459 | va_start(args, fmt); | |
460 | printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function); | |
461 | vprintk(fmt, args); | |
462 | printk("\n"); | |
463 | va_end(args); | |
464 | ||
465 | if (test_opt(sb, ERRORS_CONT)) { | |
466 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; | |
467 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
e2d67052 | 468 | ext4_commit_super(sb, 0); |
5d1b1b3f AK |
469 | return; |
470 | } | |
471 | ext4_unlock_group(sb, grp); | |
472 | ext4_handle_error(sb); | |
473 | /* | |
474 | * We only get here in the ERRORS_RO case; relocking the group | |
475 | * may be dangerous, but nothing bad will happen since the | |
476 | * filesystem will have already been marked read/only and the | |
477 | * journal has been aborted. We return 1 as a hint to callers | |
478 | * who might what to use the return value from | |
479 | * ext4_grp_locked_error() to distinguish beween the | |
480 | * ERRORS_CONT and ERRORS_RO case, and perhaps return more | |
481 | * aggressively from the ext4 function in question, with a | |
482 | * more appropriate error code. | |
483 | */ | |
484 | ext4_lock_group(sb, grp); | |
485 | return; | |
486 | } | |
487 | ||
617ba13b | 488 | void ext4_update_dynamic_rev(struct super_block *sb) |
ac27a0ec | 489 | { |
617ba13b | 490 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
ac27a0ec | 491 | |
617ba13b | 492 | if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) |
ac27a0ec DK |
493 | return; |
494 | ||
46e665e9 | 495 | ext4_warning(sb, __func__, |
ac27a0ec DK |
496 | "updating to rev %d because of new feature flag, " |
497 | "running e2fsck is recommended", | |
617ba13b | 498 | EXT4_DYNAMIC_REV); |
ac27a0ec | 499 | |
617ba13b MC |
500 | es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); |
501 | es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); | |
502 | es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); | |
ac27a0ec DK |
503 | /* leave es->s_feature_*compat flags alone */ |
504 | /* es->s_uuid will be set by e2fsck if empty */ | |
505 | ||
506 | /* | |
507 | * The rest of the superblock fields should be zero, and if not it | |
508 | * means they are likely already in use, so leave them alone. We | |
509 | * can leave it up to e2fsck to clean up any inconsistencies there. | |
510 | */ | |
511 | } | |
512 | ||
513 | /* | |
514 | * Open the external journal device | |
515 | */ | |
b31e1552 | 516 | static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) |
ac27a0ec DK |
517 | { |
518 | struct block_device *bdev; | |
519 | char b[BDEVNAME_SIZE]; | |
520 | ||
521 | bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE); | |
522 | if (IS_ERR(bdev)) | |
523 | goto fail; | |
524 | return bdev; | |
525 | ||
526 | fail: | |
b31e1552 | 527 | ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld", |
ac27a0ec DK |
528 | __bdevname(dev, b), PTR_ERR(bdev)); |
529 | return NULL; | |
530 | } | |
531 | ||
532 | /* | |
533 | * Release the journal device | |
534 | */ | |
617ba13b | 535 | static int ext4_blkdev_put(struct block_device *bdev) |
ac27a0ec DK |
536 | { |
537 | bd_release(bdev); | |
9a1c3542 | 538 | return blkdev_put(bdev, FMODE_READ|FMODE_WRITE); |
ac27a0ec DK |
539 | } |
540 | ||
617ba13b | 541 | static int ext4_blkdev_remove(struct ext4_sb_info *sbi) |
ac27a0ec DK |
542 | { |
543 | struct block_device *bdev; | |
544 | int ret = -ENODEV; | |
545 | ||
546 | bdev = sbi->journal_bdev; | |
547 | if (bdev) { | |
617ba13b | 548 | ret = ext4_blkdev_put(bdev); |
ac27a0ec DK |
549 | sbi->journal_bdev = NULL; |
550 | } | |
551 | return ret; | |
552 | } | |
553 | ||
554 | static inline struct inode *orphan_list_entry(struct list_head *l) | |
555 | { | |
617ba13b | 556 | return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; |
ac27a0ec DK |
557 | } |
558 | ||
617ba13b | 559 | static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) |
ac27a0ec DK |
560 | { |
561 | struct list_head *l; | |
562 | ||
b31e1552 ES |
563 | ext4_msg(sb, KERN_ERR, "sb orphan head is %d", |
564 | le32_to_cpu(sbi->s_es->s_last_orphan)); | |
ac27a0ec DK |
565 | |
566 | printk(KERN_ERR "sb_info orphan list:\n"); | |
567 | list_for_each(l, &sbi->s_orphan) { | |
568 | struct inode *inode = orphan_list_entry(l); | |
569 | printk(KERN_ERR " " | |
570 | "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", | |
571 | inode->i_sb->s_id, inode->i_ino, inode, | |
572 | inode->i_mode, inode->i_nlink, | |
573 | NEXT_ORPHAN(inode)); | |
574 | } | |
575 | } | |
576 | ||
2b2d6d01 | 577 | static void ext4_put_super(struct super_block *sb) |
ac27a0ec | 578 | { |
617ba13b MC |
579 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
580 | struct ext4_super_block *es = sbi->s_es; | |
ef2cabf7 | 581 | int i, err; |
ac27a0ec | 582 | |
4c0425ff MC |
583 | flush_workqueue(sbi->dio_unwritten_wq); |
584 | destroy_workqueue(sbi->dio_unwritten_wq); | |
585 | ||
a9e220f8 | 586 | lock_super(sb); |
6cfd0148 | 587 | lock_kernel(); |
8c85e125 | 588 | if (sb->s_dirt) |
ebc1ac16 | 589 | ext4_commit_super(sb, 1); |
8c85e125 | 590 | |
6fd058f7 | 591 | ext4_release_system_zone(sb); |
c9de560d | 592 | ext4_mb_release(sb); |
a86c6181 | 593 | ext4_ext_release(sb); |
617ba13b | 594 | ext4_xattr_put_super(sb); |
0390131b FM |
595 | if (sbi->s_journal) { |
596 | err = jbd2_journal_destroy(sbi->s_journal); | |
597 | sbi->s_journal = NULL; | |
598 | if (err < 0) | |
599 | ext4_abort(sb, __func__, | |
600 | "Couldn't clean up the journal"); | |
601 | } | |
ac27a0ec | 602 | if (!(sb->s_flags & MS_RDONLY)) { |
617ba13b | 603 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); |
ac27a0ec | 604 | es->s_state = cpu_to_le16(sbi->s_mount_state); |
e2d67052 | 605 | ext4_commit_super(sb, 1); |
ac27a0ec | 606 | } |
240799cd | 607 | if (sbi->s_proc) { |
9f6200bb | 608 | remove_proc_entry(sb->s_id, ext4_proc_root); |
240799cd | 609 | } |
3197ebdb | 610 | kobject_del(&sbi->s_kobj); |
ac27a0ec DK |
611 | |
612 | for (i = 0; i < sbi->s_gdb_count; i++) | |
613 | brelse(sbi->s_group_desc[i]); | |
614 | kfree(sbi->s_group_desc); | |
c5ca7c76 TT |
615 | if (is_vmalloc_addr(sbi->s_flex_groups)) |
616 | vfree(sbi->s_flex_groups); | |
617 | else | |
618 | kfree(sbi->s_flex_groups); | |
ac27a0ec DK |
619 | percpu_counter_destroy(&sbi->s_freeblocks_counter); |
620 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
621 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
6bc6e63f | 622 | percpu_counter_destroy(&sbi->s_dirtyblocks_counter); |
ac27a0ec DK |
623 | brelse(sbi->s_sbh); |
624 | #ifdef CONFIG_QUOTA | |
625 | for (i = 0; i < MAXQUOTAS; i++) | |
626 | kfree(sbi->s_qf_names[i]); | |
627 | #endif | |
628 | ||
629 | /* Debugging code just in case the in-memory inode orphan list | |
630 | * isn't empty. The on-disk one can be non-empty if we've | |
631 | * detected an error and taken the fs readonly, but the | |
632 | * in-memory list had better be clean by this point. */ | |
633 | if (!list_empty(&sbi->s_orphan)) | |
634 | dump_orphan_list(sb, sbi); | |
635 | J_ASSERT(list_empty(&sbi->s_orphan)); | |
636 | ||
f98393a6 | 637 | invalidate_bdev(sb->s_bdev); |
ac27a0ec DK |
638 | if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { |
639 | /* | |
640 | * Invalidate the journal device's buffers. We don't want them | |
641 | * floating about in memory - the physical journal device may | |
642 | * hotswapped, and it breaks the `ro-after' testing code. | |
643 | */ | |
644 | sync_blockdev(sbi->journal_bdev); | |
f98393a6 | 645 | invalidate_bdev(sbi->journal_bdev); |
617ba13b | 646 | ext4_blkdev_remove(sbi); |
ac27a0ec DK |
647 | } |
648 | sb->s_fs_info = NULL; | |
3197ebdb TT |
649 | /* |
650 | * Now that we are completely done shutting down the | |
651 | * superblock, we need to actually destroy the kobject. | |
652 | */ | |
653 | unlock_kernel(); | |
654 | unlock_super(sb); | |
655 | kobject_put(&sbi->s_kobj); | |
656 | wait_for_completion(&sbi->s_kobj_unregister); | |
705895b6 | 657 | kfree(sbi->s_blockgroup_lock); |
ac27a0ec | 658 | kfree(sbi); |
ac27a0ec DK |
659 | } |
660 | ||
e18b890b | 661 | static struct kmem_cache *ext4_inode_cachep; |
ac27a0ec DK |
662 | |
663 | /* | |
664 | * Called inside transaction, so use GFP_NOFS | |
665 | */ | |
617ba13b | 666 | static struct inode *ext4_alloc_inode(struct super_block *sb) |
ac27a0ec | 667 | { |
617ba13b | 668 | struct ext4_inode_info *ei; |
ac27a0ec | 669 | |
e6b4f8da | 670 | ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); |
ac27a0ec DK |
671 | if (!ei) |
672 | return NULL; | |
0b8e58a1 | 673 | |
ac27a0ec | 674 | ei->vfs_inode.i_version = 1; |
91246c00 | 675 | ei->vfs_inode.i_data.writeback_index = 0; |
a86c6181 | 676 | memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache)); |
c9de560d AT |
677 | INIT_LIST_HEAD(&ei->i_prealloc_list); |
678 | spin_lock_init(&ei->i_prealloc_lock); | |
0390131b FM |
679 | /* |
680 | * Note: We can be called before EXT4_SB(sb)->s_journal is set, | |
681 | * therefore it can be null here. Don't check it, just initialize | |
682 | * jinode. | |
683 | */ | |
678aaf48 | 684 | jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode); |
d2a17637 MC |
685 | ei->i_reserved_data_blocks = 0; |
686 | ei->i_reserved_meta_blocks = 0; | |
687 | ei->i_allocated_meta_blocks = 0; | |
688 | ei->i_delalloc_reserved_flag = 0; | |
689 | spin_lock_init(&(ei->i_block_reservation_lock)); | |
8d5d02e6 MC |
690 | INIT_LIST_HEAD(&ei->i_aio_dio_complete_list); |
691 | ei->cur_aio_dio = NULL; | |
0b8e58a1 | 692 | |
ac27a0ec DK |
693 | return &ei->vfs_inode; |
694 | } | |
695 | ||
617ba13b | 696 | static void ext4_destroy_inode(struct inode *inode) |
ac27a0ec | 697 | { |
9f7dd93d | 698 | if (!list_empty(&(EXT4_I(inode)->i_orphan))) { |
b31e1552 ES |
699 | ext4_msg(inode->i_sb, KERN_ERR, |
700 | "Inode %lu (%p): orphan list check failed!", | |
701 | inode->i_ino, EXT4_I(inode)); | |
9f7dd93d VA |
702 | print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, |
703 | EXT4_I(inode), sizeof(struct ext4_inode_info), | |
704 | true); | |
705 | dump_stack(); | |
706 | } | |
617ba13b | 707 | kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); |
ac27a0ec DK |
708 | } |
709 | ||
51cc5068 | 710 | static void init_once(void *foo) |
ac27a0ec | 711 | { |
617ba13b | 712 | struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; |
ac27a0ec | 713 | |
a35afb83 | 714 | INIT_LIST_HEAD(&ei->i_orphan); |
03010a33 | 715 | #ifdef CONFIG_EXT4_FS_XATTR |
a35afb83 | 716 | init_rwsem(&ei->xattr_sem); |
ac27a0ec | 717 | #endif |
0e855ac8 | 718 | init_rwsem(&ei->i_data_sem); |
a35afb83 | 719 | inode_init_once(&ei->vfs_inode); |
ac27a0ec DK |
720 | } |
721 | ||
722 | static int init_inodecache(void) | |
723 | { | |
617ba13b MC |
724 | ext4_inode_cachep = kmem_cache_create("ext4_inode_cache", |
725 | sizeof(struct ext4_inode_info), | |
ac27a0ec DK |
726 | 0, (SLAB_RECLAIM_ACCOUNT| |
727 | SLAB_MEM_SPREAD), | |
20c2df83 | 728 | init_once); |
617ba13b | 729 | if (ext4_inode_cachep == NULL) |
ac27a0ec DK |
730 | return -ENOMEM; |
731 | return 0; | |
732 | } | |
733 | ||
734 | static void destroy_inodecache(void) | |
735 | { | |
617ba13b | 736 | kmem_cache_destroy(ext4_inode_cachep); |
ac27a0ec DK |
737 | } |
738 | ||
617ba13b | 739 | static void ext4_clear_inode(struct inode *inode) |
ac27a0ec | 740 | { |
c2ea3fde | 741 | ext4_discard_preallocations(inode); |
0390131b FM |
742 | if (EXT4_JOURNAL(inode)) |
743 | jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal, | |
678aaf48 | 744 | &EXT4_I(inode)->jinode); |
ac27a0ec DK |
745 | } |
746 | ||
2b2d6d01 TT |
747 | static inline void ext4_show_quota_options(struct seq_file *seq, |
748 | struct super_block *sb) | |
ac27a0ec DK |
749 | { |
750 | #if defined(CONFIG_QUOTA) | |
617ba13b | 751 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec DK |
752 | |
753 | if (sbi->s_jquota_fmt) | |
754 | seq_printf(seq, ",jqfmt=%s", | |
af5bc92d | 755 | (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0"); |
ac27a0ec DK |
756 | |
757 | if (sbi->s_qf_names[USRQUOTA]) | |
758 | seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]); | |
759 | ||
760 | if (sbi->s_qf_names[GRPQUOTA]) | |
761 | seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]); | |
762 | ||
617ba13b | 763 | if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) |
ac27a0ec DK |
764 | seq_puts(seq, ",usrquota"); |
765 | ||
617ba13b | 766 | if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) |
ac27a0ec DK |
767 | seq_puts(seq, ",grpquota"); |
768 | #endif | |
769 | } | |
770 | ||
d9c9bef1 MS |
771 | /* |
772 | * Show an option if | |
773 | * - it's set to a non-default value OR | |
774 | * - if the per-sb default is different from the global default | |
775 | */ | |
617ba13b | 776 | static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs) |
ac27a0ec | 777 | { |
aa22df2c AK |
778 | int def_errors; |
779 | unsigned long def_mount_opts; | |
ac27a0ec | 780 | struct super_block *sb = vfs->mnt_sb; |
d9c9bef1 MS |
781 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
782 | struct ext4_super_block *es = sbi->s_es; | |
d9c9bef1 MS |
783 | |
784 | def_mount_opts = le32_to_cpu(es->s_default_mount_opts); | |
aa22df2c | 785 | def_errors = le16_to_cpu(es->s_errors); |
d9c9bef1 MS |
786 | |
787 | if (sbi->s_sb_block != 1) | |
788 | seq_printf(seq, ",sb=%llu", sbi->s_sb_block); | |
789 | if (test_opt(sb, MINIX_DF)) | |
790 | seq_puts(seq, ",minixdf"); | |
aa22df2c | 791 | if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS)) |
d9c9bef1 MS |
792 | seq_puts(seq, ",grpid"); |
793 | if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS)) | |
794 | seq_puts(seq, ",nogrpid"); | |
795 | if (sbi->s_resuid != EXT4_DEF_RESUID || | |
796 | le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) { | |
797 | seq_printf(seq, ",resuid=%u", sbi->s_resuid); | |
798 | } | |
799 | if (sbi->s_resgid != EXT4_DEF_RESGID || | |
800 | le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) { | |
801 | seq_printf(seq, ",resgid=%u", sbi->s_resgid); | |
802 | } | |
bb4f397a | 803 | if (test_opt(sb, ERRORS_RO)) { |
d9c9bef1 | 804 | if (def_errors == EXT4_ERRORS_PANIC || |
bb4f397a AK |
805 | def_errors == EXT4_ERRORS_CONTINUE) { |
806 | seq_puts(seq, ",errors=remount-ro"); | |
d9c9bef1 MS |
807 | } |
808 | } | |
aa22df2c | 809 | if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) |
bb4f397a | 810 | seq_puts(seq, ",errors=continue"); |
aa22df2c | 811 | if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) |
d9c9bef1 | 812 | seq_puts(seq, ",errors=panic"); |
aa22df2c | 813 | if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16)) |
d9c9bef1 | 814 | seq_puts(seq, ",nouid32"); |
aa22df2c | 815 | if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG)) |
d9c9bef1 MS |
816 | seq_puts(seq, ",debug"); |
817 | if (test_opt(sb, OLDALLOC)) | |
818 | seq_puts(seq, ",oldalloc"); | |
03010a33 | 819 | #ifdef CONFIG_EXT4_FS_XATTR |
aa22df2c AK |
820 | if (test_opt(sb, XATTR_USER) && |
821 | !(def_mount_opts & EXT4_DEFM_XATTR_USER)) | |
d9c9bef1 MS |
822 | seq_puts(seq, ",user_xattr"); |
823 | if (!test_opt(sb, XATTR_USER) && | |
824 | (def_mount_opts & EXT4_DEFM_XATTR_USER)) { | |
825 | seq_puts(seq, ",nouser_xattr"); | |
826 | } | |
827 | #endif | |
03010a33 | 828 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
aa22df2c | 829 | if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL)) |
d9c9bef1 MS |
830 | seq_puts(seq, ",acl"); |
831 | if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL)) | |
832 | seq_puts(seq, ",noacl"); | |
833 | #endif | |
30773840 | 834 | if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { |
d9c9bef1 MS |
835 | seq_printf(seq, ",commit=%u", |
836 | (unsigned) (sbi->s_commit_interval / HZ)); | |
837 | } | |
30773840 TT |
838 | if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) { |
839 | seq_printf(seq, ",min_batch_time=%u", | |
840 | (unsigned) sbi->s_min_batch_time); | |
841 | } | |
842 | if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) { | |
843 | seq_printf(seq, ",max_batch_time=%u", | |
844 | (unsigned) sbi->s_min_batch_time); | |
845 | } | |
846 | ||
571640ca ES |
847 | /* |
848 | * We're changing the default of barrier mount option, so | |
849 | * let's always display its mount state so it's clear what its | |
850 | * status is. | |
851 | */ | |
852 | seq_puts(seq, ",barrier="); | |
853 | seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0"); | |
cd0b6a39 TT |
854 | if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) |
855 | seq_puts(seq, ",journal_async_commit"); | |
d9c9bef1 MS |
856 | if (test_opt(sb, NOBH)) |
857 | seq_puts(seq, ",nobh"); | |
25ec56b5 JNC |
858 | if (test_opt(sb, I_VERSION)) |
859 | seq_puts(seq, ",i_version"); | |
dd919b98 AK |
860 | if (!test_opt(sb, DELALLOC)) |
861 | seq_puts(seq, ",nodelalloc"); | |
862 | ||
ac27a0ec | 863 | |
cb45bbe4 MS |
864 | if (sbi->s_stripe) |
865 | seq_printf(seq, ",stripe=%lu", sbi->s_stripe); | |
aa22df2c AK |
866 | /* |
867 | * journal mode get enabled in different ways | |
868 | * So just print the value even if we didn't specify it | |
869 | */ | |
617ba13b | 870 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) |
ac27a0ec | 871 | seq_puts(seq, ",data=journal"); |
617ba13b | 872 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) |
ac27a0ec | 873 | seq_puts(seq, ",data=ordered"); |
617ba13b | 874 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) |
ac27a0ec DK |
875 | seq_puts(seq, ",data=writeback"); |
876 | ||
240799cd TT |
877 | if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) |
878 | seq_printf(seq, ",inode_readahead_blks=%u", | |
879 | sbi->s_inode_readahead_blks); | |
880 | ||
5bf5683a HK |
881 | if (test_opt(sb, DATA_ERR_ABORT)) |
882 | seq_puts(seq, ",data_err=abort"); | |
883 | ||
afd4672d | 884 | if (test_opt(sb, NO_AUTO_DA_ALLOC)) |
06705bff | 885 | seq_puts(seq, ",noauto_da_alloc"); |
afd4672d | 886 | |
617ba13b | 887 | ext4_show_quota_options(seq, sb); |
0b8e58a1 | 888 | |
ac27a0ec DK |
889 | return 0; |
890 | } | |
891 | ||
1b961ac0 | 892 | static struct inode *ext4_nfs_get_inode(struct super_block *sb, |
0b8e58a1 | 893 | u64 ino, u32 generation) |
ac27a0ec | 894 | { |
ac27a0ec | 895 | struct inode *inode; |
ac27a0ec | 896 | |
617ba13b | 897 | if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO) |
ac27a0ec | 898 | return ERR_PTR(-ESTALE); |
617ba13b | 899 | if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count)) |
ac27a0ec DK |
900 | return ERR_PTR(-ESTALE); |
901 | ||
902 | /* iget isn't really right if the inode is currently unallocated!! | |
903 | * | |
617ba13b | 904 | * ext4_read_inode will return a bad_inode if the inode had been |
ac27a0ec DK |
905 | * deleted, so we should be safe. |
906 | * | |
907 | * Currently we don't know the generation for parent directory, so | |
908 | * a generation of 0 means "accept any" | |
909 | */ | |
1d1fe1ee DH |
910 | inode = ext4_iget(sb, ino); |
911 | if (IS_ERR(inode)) | |
912 | return ERR_CAST(inode); | |
913 | if (generation && inode->i_generation != generation) { | |
ac27a0ec DK |
914 | iput(inode); |
915 | return ERR_PTR(-ESTALE); | |
916 | } | |
1b961ac0 CH |
917 | |
918 | return inode; | |
919 | } | |
920 | ||
921 | static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, | |
0b8e58a1 | 922 | int fh_len, int fh_type) |
1b961ac0 CH |
923 | { |
924 | return generic_fh_to_dentry(sb, fid, fh_len, fh_type, | |
925 | ext4_nfs_get_inode); | |
926 | } | |
927 | ||
928 | static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, | |
0b8e58a1 | 929 | int fh_len, int fh_type) |
1b961ac0 CH |
930 | { |
931 | return generic_fh_to_parent(sb, fid, fh_len, fh_type, | |
932 | ext4_nfs_get_inode); | |
ac27a0ec DK |
933 | } |
934 | ||
c39a7f84 TO |
935 | /* |
936 | * Try to release metadata pages (indirect blocks, directories) which are | |
937 | * mapped via the block device. Since these pages could have journal heads | |
938 | * which would prevent try_to_free_buffers() from freeing them, we must use | |
939 | * jbd2 layer's try_to_free_buffers() function to release them. | |
940 | */ | |
0b8e58a1 AD |
941 | static int bdev_try_to_free_page(struct super_block *sb, struct page *page, |
942 | gfp_t wait) | |
c39a7f84 TO |
943 | { |
944 | journal_t *journal = EXT4_SB(sb)->s_journal; | |
945 | ||
946 | WARN_ON(PageChecked(page)); | |
947 | if (!page_has_buffers(page)) | |
948 | return 0; | |
949 | if (journal) | |
950 | return jbd2_journal_try_to_free_buffers(journal, page, | |
951 | wait & ~__GFP_WAIT); | |
952 | return try_to_free_buffers(page); | |
953 | } | |
954 | ||
ac27a0ec | 955 | #ifdef CONFIG_QUOTA |
af5bc92d | 956 | #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group") |
2b2d6d01 | 957 | #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA)) |
ac27a0ec | 958 | |
617ba13b MC |
959 | static int ext4_write_dquot(struct dquot *dquot); |
960 | static int ext4_acquire_dquot(struct dquot *dquot); | |
961 | static int ext4_release_dquot(struct dquot *dquot); | |
962 | static int ext4_mark_dquot_dirty(struct dquot *dquot); | |
963 | static int ext4_write_info(struct super_block *sb, int type); | |
6f28e087 JK |
964 | static int ext4_quota_on(struct super_block *sb, int type, int format_id, |
965 | char *path, int remount); | |
617ba13b MC |
966 | static int ext4_quota_on_mount(struct super_block *sb, int type); |
967 | static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, | |
ac27a0ec | 968 | size_t len, loff_t off); |
617ba13b | 969 | static ssize_t ext4_quota_write(struct super_block *sb, int type, |
ac27a0ec DK |
970 | const char *data, size_t len, loff_t off); |
971 | ||
61e225dc | 972 | static const struct dquot_operations ext4_quota_operations = { |
edf72453 JK |
973 | .initialize = dquot_initialize, |
974 | .drop = dquot_drop, | |
ac27a0ec | 975 | .alloc_space = dquot_alloc_space, |
60e58e0f MC |
976 | .reserve_space = dquot_reserve_space, |
977 | .claim_space = dquot_claim_space, | |
978 | .release_rsv = dquot_release_reserved_space, | |
979 | .get_reserved_space = ext4_get_reserved_space, | |
ac27a0ec DK |
980 | .alloc_inode = dquot_alloc_inode, |
981 | .free_space = dquot_free_space, | |
982 | .free_inode = dquot_free_inode, | |
983 | .transfer = dquot_transfer, | |
617ba13b MC |
984 | .write_dquot = ext4_write_dquot, |
985 | .acquire_dquot = ext4_acquire_dquot, | |
986 | .release_dquot = ext4_release_dquot, | |
987 | .mark_dirty = ext4_mark_dquot_dirty, | |
a5b5ee32 JK |
988 | .write_info = ext4_write_info, |
989 | .alloc_dquot = dquot_alloc, | |
990 | .destroy_dquot = dquot_destroy, | |
ac27a0ec DK |
991 | }; |
992 | ||
0d54b217 | 993 | static const struct quotactl_ops ext4_qctl_operations = { |
617ba13b | 994 | .quota_on = ext4_quota_on, |
ac27a0ec DK |
995 | .quota_off = vfs_quota_off, |
996 | .quota_sync = vfs_quota_sync, | |
997 | .get_info = vfs_get_dqinfo, | |
998 | .set_info = vfs_set_dqinfo, | |
999 | .get_dqblk = vfs_get_dqblk, | |
1000 | .set_dqblk = vfs_set_dqblk | |
1001 | }; | |
1002 | #endif | |
1003 | ||
ee9b6d61 | 1004 | static const struct super_operations ext4_sops = { |
617ba13b MC |
1005 | .alloc_inode = ext4_alloc_inode, |
1006 | .destroy_inode = ext4_destroy_inode, | |
617ba13b MC |
1007 | .write_inode = ext4_write_inode, |
1008 | .dirty_inode = ext4_dirty_inode, | |
1009 | .delete_inode = ext4_delete_inode, | |
1010 | .put_super = ext4_put_super, | |
617ba13b | 1011 | .sync_fs = ext4_sync_fs, |
c4be0c1d TS |
1012 | .freeze_fs = ext4_freeze, |
1013 | .unfreeze_fs = ext4_unfreeze, | |
617ba13b MC |
1014 | .statfs = ext4_statfs, |
1015 | .remount_fs = ext4_remount, | |
1016 | .clear_inode = ext4_clear_inode, | |
1017 | .show_options = ext4_show_options, | |
ac27a0ec | 1018 | #ifdef CONFIG_QUOTA |
617ba13b MC |
1019 | .quota_read = ext4_quota_read, |
1020 | .quota_write = ext4_quota_write, | |
ac27a0ec | 1021 | #endif |
c39a7f84 | 1022 | .bdev_try_to_free_page = bdev_try_to_free_page, |
ac27a0ec DK |
1023 | }; |
1024 | ||
9ca92389 TT |
1025 | static const struct super_operations ext4_nojournal_sops = { |
1026 | .alloc_inode = ext4_alloc_inode, | |
1027 | .destroy_inode = ext4_destroy_inode, | |
1028 | .write_inode = ext4_write_inode, | |
1029 | .dirty_inode = ext4_dirty_inode, | |
1030 | .delete_inode = ext4_delete_inode, | |
1031 | .write_super = ext4_write_super, | |
1032 | .put_super = ext4_put_super, | |
1033 | .statfs = ext4_statfs, | |
1034 | .remount_fs = ext4_remount, | |
1035 | .clear_inode = ext4_clear_inode, | |
1036 | .show_options = ext4_show_options, | |
1037 | #ifdef CONFIG_QUOTA | |
1038 | .quota_read = ext4_quota_read, | |
1039 | .quota_write = ext4_quota_write, | |
1040 | #endif | |
1041 | .bdev_try_to_free_page = bdev_try_to_free_page, | |
1042 | }; | |
1043 | ||
39655164 | 1044 | static const struct export_operations ext4_export_ops = { |
1b961ac0 CH |
1045 | .fh_to_dentry = ext4_fh_to_dentry, |
1046 | .fh_to_parent = ext4_fh_to_parent, | |
617ba13b | 1047 | .get_parent = ext4_get_parent, |
ac27a0ec DK |
1048 | }; |
1049 | ||
1050 | enum { | |
1051 | Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, | |
1052 | Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, | |
01436ef2 | 1053 | Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov, |
ac27a0ec | 1054 | Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, |
06705bff | 1055 | Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh, |
30773840 | 1056 | Opt_commit, Opt_min_batch_time, Opt_max_batch_time, |
c3191067 | 1057 | Opt_journal_update, Opt_journal_dev, |
818d276c | 1058 | Opt_journal_checksum, Opt_journal_async_commit, |
ac27a0ec | 1059 | Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, |
f4033903 | 1060 | Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length, |
ac27a0ec DK |
1061 | Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, |
1062 | Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota, | |
06705bff TT |
1063 | Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize, |
1064 | Opt_usrquota, Opt_grpquota, Opt_i_version, | |
01436ef2 | 1065 | Opt_stripe, Opt_delalloc, Opt_nodelalloc, |
6fd058f7 | 1066 | Opt_block_validity, Opt_noblock_validity, |
b3881f74 | 1067 | Opt_inode_readahead_blks, Opt_journal_ioprio |
ac27a0ec DK |
1068 | }; |
1069 | ||
a447c093 | 1070 | static const match_table_t tokens = { |
ac27a0ec DK |
1071 | {Opt_bsd_df, "bsddf"}, |
1072 | {Opt_minix_df, "minixdf"}, | |
1073 | {Opt_grpid, "grpid"}, | |
1074 | {Opt_grpid, "bsdgroups"}, | |
1075 | {Opt_nogrpid, "nogrpid"}, | |
1076 | {Opt_nogrpid, "sysvgroups"}, | |
1077 | {Opt_resgid, "resgid=%u"}, | |
1078 | {Opt_resuid, "resuid=%u"}, | |
1079 | {Opt_sb, "sb=%u"}, | |
1080 | {Opt_err_cont, "errors=continue"}, | |
1081 | {Opt_err_panic, "errors=panic"}, | |
1082 | {Opt_err_ro, "errors=remount-ro"}, | |
1083 | {Opt_nouid32, "nouid32"}, | |
ac27a0ec DK |
1084 | {Opt_debug, "debug"}, |
1085 | {Opt_oldalloc, "oldalloc"}, | |
1086 | {Opt_orlov, "orlov"}, | |
1087 | {Opt_user_xattr, "user_xattr"}, | |
1088 | {Opt_nouser_xattr, "nouser_xattr"}, | |
1089 | {Opt_acl, "acl"}, | |
1090 | {Opt_noacl, "noacl"}, | |
ac27a0ec DK |
1091 | {Opt_noload, "noload"}, |
1092 | {Opt_nobh, "nobh"}, | |
1093 | {Opt_bh, "bh"}, | |
1094 | {Opt_commit, "commit=%u"}, | |
30773840 TT |
1095 | {Opt_min_batch_time, "min_batch_time=%u"}, |
1096 | {Opt_max_batch_time, "max_batch_time=%u"}, | |
ac27a0ec | 1097 | {Opt_journal_update, "journal=update"}, |
ac27a0ec | 1098 | {Opt_journal_dev, "journal_dev=%u"}, |
818d276c GS |
1099 | {Opt_journal_checksum, "journal_checksum"}, |
1100 | {Opt_journal_async_commit, "journal_async_commit"}, | |
ac27a0ec DK |
1101 | {Opt_abort, "abort"}, |
1102 | {Opt_data_journal, "data=journal"}, | |
1103 | {Opt_data_ordered, "data=ordered"}, | |
1104 | {Opt_data_writeback, "data=writeback"}, | |
5bf5683a HK |
1105 | {Opt_data_err_abort, "data_err=abort"}, |
1106 | {Opt_data_err_ignore, "data_err=ignore"}, | |
f4033903 | 1107 | {Opt_mb_history_length, "mb_history_length=%u"}, |
ac27a0ec DK |
1108 | {Opt_offusrjquota, "usrjquota="}, |
1109 | {Opt_usrjquota, "usrjquota=%s"}, | |
1110 | {Opt_offgrpjquota, "grpjquota="}, | |
1111 | {Opt_grpjquota, "grpjquota=%s"}, | |
1112 | {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, | |
1113 | {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, | |
1114 | {Opt_grpquota, "grpquota"}, | |
1115 | {Opt_noquota, "noquota"}, | |
1116 | {Opt_quota, "quota"}, | |
1117 | {Opt_usrquota, "usrquota"}, | |
1118 | {Opt_barrier, "barrier=%u"}, | |
06705bff TT |
1119 | {Opt_barrier, "barrier"}, |
1120 | {Opt_nobarrier, "nobarrier"}, | |
25ec56b5 | 1121 | {Opt_i_version, "i_version"}, |
c9de560d | 1122 | {Opt_stripe, "stripe=%u"}, |
ac27a0ec | 1123 | {Opt_resize, "resize"}, |
64769240 | 1124 | {Opt_delalloc, "delalloc"}, |
dd919b98 | 1125 | {Opt_nodelalloc, "nodelalloc"}, |
6fd058f7 TT |
1126 | {Opt_block_validity, "block_validity"}, |
1127 | {Opt_noblock_validity, "noblock_validity"}, | |
240799cd | 1128 | {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, |
b3881f74 | 1129 | {Opt_journal_ioprio, "journal_ioprio=%u"}, |
afd4672d | 1130 | {Opt_auto_da_alloc, "auto_da_alloc=%u"}, |
06705bff TT |
1131 | {Opt_auto_da_alloc, "auto_da_alloc"}, |
1132 | {Opt_noauto_da_alloc, "noauto_da_alloc"}, | |
f3f12faa | 1133 | {Opt_err, NULL}, |
ac27a0ec DK |
1134 | }; |
1135 | ||
617ba13b | 1136 | static ext4_fsblk_t get_sb_block(void **data) |
ac27a0ec | 1137 | { |
617ba13b | 1138 | ext4_fsblk_t sb_block; |
ac27a0ec DK |
1139 | char *options = (char *) *data; |
1140 | ||
1141 | if (!options || strncmp(options, "sb=", 3) != 0) | |
1142 | return 1; /* Default location */ | |
0b8e58a1 | 1143 | |
ac27a0ec | 1144 | options += 3; |
0b8e58a1 | 1145 | /* TODO: use simple_strtoll with >32bit ext4 */ |
ac27a0ec DK |
1146 | sb_block = simple_strtoul(options, &options, 0); |
1147 | if (*options && *options != ',') { | |
4776004f | 1148 | printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", |
ac27a0ec DK |
1149 | (char *) *data); |
1150 | return 1; | |
1151 | } | |
1152 | if (*options == ',') | |
1153 | options++; | |
1154 | *data = (void *) options; | |
0b8e58a1 | 1155 | |
ac27a0ec DK |
1156 | return sb_block; |
1157 | } | |
1158 | ||
b3881f74 TT |
1159 | #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) |
1160 | ||
2b2d6d01 | 1161 | static int parse_options(char *options, struct super_block *sb, |
c3191067 | 1162 | unsigned long *journal_devnum, |
b3881f74 | 1163 | unsigned int *journal_ioprio, |
2b2d6d01 | 1164 | ext4_fsblk_t *n_blocks_count, int is_remount) |
ac27a0ec | 1165 | { |
617ba13b | 1166 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
2b2d6d01 | 1167 | char *p; |
ac27a0ec DK |
1168 | substring_t args[MAX_OPT_ARGS]; |
1169 | int data_opt = 0; | |
1170 | int option; | |
1171 | #ifdef CONFIG_QUOTA | |
dfc5d03f | 1172 | int qtype, qfmt; |
ac27a0ec DK |
1173 | char *qname; |
1174 | #endif | |
1175 | ||
1176 | if (!options) | |
1177 | return 1; | |
1178 | ||
2b2d6d01 | 1179 | while ((p = strsep(&options, ",")) != NULL) { |
ac27a0ec DK |
1180 | int token; |
1181 | if (!*p) | |
1182 | continue; | |
1183 | ||
1184 | token = match_token(p, tokens, args); | |
1185 | switch (token) { | |
1186 | case Opt_bsd_df: | |
2b2d6d01 | 1187 | clear_opt(sbi->s_mount_opt, MINIX_DF); |
ac27a0ec DK |
1188 | break; |
1189 | case Opt_minix_df: | |
2b2d6d01 | 1190 | set_opt(sbi->s_mount_opt, MINIX_DF); |
ac27a0ec DK |
1191 | break; |
1192 | case Opt_grpid: | |
2b2d6d01 | 1193 | set_opt(sbi->s_mount_opt, GRPID); |
ac27a0ec DK |
1194 | break; |
1195 | case Opt_nogrpid: | |
2b2d6d01 | 1196 | clear_opt(sbi->s_mount_opt, GRPID); |
ac27a0ec DK |
1197 | break; |
1198 | case Opt_resuid: | |
1199 | if (match_int(&args[0], &option)) | |
1200 | return 0; | |
1201 | sbi->s_resuid = option; | |
1202 | break; | |
1203 | case Opt_resgid: | |
1204 | if (match_int(&args[0], &option)) | |
1205 | return 0; | |
1206 | sbi->s_resgid = option; | |
1207 | break; | |
1208 | case Opt_sb: | |
1209 | /* handled by get_sb_block() instead of here */ | |
1210 | /* *sb_block = match_int(&args[0]); */ | |
1211 | break; | |
1212 | case Opt_err_panic: | |
2b2d6d01 TT |
1213 | clear_opt(sbi->s_mount_opt, ERRORS_CONT); |
1214 | clear_opt(sbi->s_mount_opt, ERRORS_RO); | |
1215 | set_opt(sbi->s_mount_opt, ERRORS_PANIC); | |
ac27a0ec DK |
1216 | break; |
1217 | case Opt_err_ro: | |
2b2d6d01 TT |
1218 | clear_opt(sbi->s_mount_opt, ERRORS_CONT); |
1219 | clear_opt(sbi->s_mount_opt, ERRORS_PANIC); | |
1220 | set_opt(sbi->s_mount_opt, ERRORS_RO); | |
ac27a0ec DK |
1221 | break; |
1222 | case Opt_err_cont: | |
2b2d6d01 TT |
1223 | clear_opt(sbi->s_mount_opt, ERRORS_RO); |
1224 | clear_opt(sbi->s_mount_opt, ERRORS_PANIC); | |
1225 | set_opt(sbi->s_mount_opt, ERRORS_CONT); | |
ac27a0ec DK |
1226 | break; |
1227 | case Opt_nouid32: | |
2b2d6d01 | 1228 | set_opt(sbi->s_mount_opt, NO_UID32); |
ac27a0ec | 1229 | break; |
ac27a0ec | 1230 | case Opt_debug: |
2b2d6d01 | 1231 | set_opt(sbi->s_mount_opt, DEBUG); |
ac27a0ec DK |
1232 | break; |
1233 | case Opt_oldalloc: | |
2b2d6d01 | 1234 | set_opt(sbi->s_mount_opt, OLDALLOC); |
ac27a0ec DK |
1235 | break; |
1236 | case Opt_orlov: | |
2b2d6d01 | 1237 | clear_opt(sbi->s_mount_opt, OLDALLOC); |
ac27a0ec | 1238 | break; |
03010a33 | 1239 | #ifdef CONFIG_EXT4_FS_XATTR |
ac27a0ec | 1240 | case Opt_user_xattr: |
2b2d6d01 | 1241 | set_opt(sbi->s_mount_opt, XATTR_USER); |
ac27a0ec DK |
1242 | break; |
1243 | case Opt_nouser_xattr: | |
2b2d6d01 | 1244 | clear_opt(sbi->s_mount_opt, XATTR_USER); |
ac27a0ec DK |
1245 | break; |
1246 | #else | |
1247 | case Opt_user_xattr: | |
1248 | case Opt_nouser_xattr: | |
b31e1552 | 1249 | ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported"); |
ac27a0ec DK |
1250 | break; |
1251 | #endif | |
03010a33 | 1252 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
ac27a0ec DK |
1253 | case Opt_acl: |
1254 | set_opt(sbi->s_mount_opt, POSIX_ACL); | |
1255 | break; | |
1256 | case Opt_noacl: | |
1257 | clear_opt(sbi->s_mount_opt, POSIX_ACL); | |
1258 | break; | |
1259 | #else | |
1260 | case Opt_acl: | |
1261 | case Opt_noacl: | |
b31e1552 | 1262 | ext4_msg(sb, KERN_ERR, "(no)acl options not supported"); |
ac27a0ec DK |
1263 | break; |
1264 | #endif | |
ac27a0ec DK |
1265 | case Opt_journal_update: |
1266 | /* @@@ FIXME */ | |
1267 | /* Eventually we will want to be able to create | |
1268 | a journal file here. For now, only allow the | |
1269 | user to specify an existing inode to be the | |
1270 | journal file. */ | |
1271 | if (is_remount) { | |
b31e1552 ES |
1272 | ext4_msg(sb, KERN_ERR, |
1273 | "Cannot specify journal on remount"); | |
ac27a0ec DK |
1274 | return 0; |
1275 | } | |
2b2d6d01 | 1276 | set_opt(sbi->s_mount_opt, UPDATE_JOURNAL); |
ac27a0ec | 1277 | break; |
ac27a0ec DK |
1278 | case Opt_journal_dev: |
1279 | if (is_remount) { | |
b31e1552 ES |
1280 | ext4_msg(sb, KERN_ERR, |
1281 | "Cannot specify journal on remount"); | |
ac27a0ec DK |
1282 | return 0; |
1283 | } | |
1284 | if (match_int(&args[0], &option)) | |
1285 | return 0; | |
1286 | *journal_devnum = option; | |
1287 | break; | |
818d276c | 1288 | case Opt_journal_checksum: |
d0646f7b | 1289 | break; /* Kept for backwards compatibility */ |
818d276c GS |
1290 | case Opt_journal_async_commit: |
1291 | set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT); | |
818d276c | 1292 | break; |
ac27a0ec | 1293 | case Opt_noload: |
2b2d6d01 | 1294 | set_opt(sbi->s_mount_opt, NOLOAD); |
ac27a0ec DK |
1295 | break; |
1296 | case Opt_commit: | |
1297 | if (match_int(&args[0], &option)) | |
1298 | return 0; | |
1299 | if (option < 0) | |
1300 | return 0; | |
1301 | if (option == 0) | |
cd02ff0b | 1302 | option = JBD2_DEFAULT_MAX_COMMIT_AGE; |
ac27a0ec DK |
1303 | sbi->s_commit_interval = HZ * option; |
1304 | break; | |
30773840 TT |
1305 | case Opt_max_batch_time: |
1306 | if (match_int(&args[0], &option)) | |
1307 | return 0; | |
1308 | if (option < 0) | |
1309 | return 0; | |
1310 | if (option == 0) | |
1311 | option = EXT4_DEF_MAX_BATCH_TIME; | |
1312 | sbi->s_max_batch_time = option; | |
1313 | break; | |
1314 | case Opt_min_batch_time: | |
1315 | if (match_int(&args[0], &option)) | |
1316 | return 0; | |
1317 | if (option < 0) | |
1318 | return 0; | |
1319 | sbi->s_min_batch_time = option; | |
1320 | break; | |
ac27a0ec | 1321 | case Opt_data_journal: |
617ba13b | 1322 | data_opt = EXT4_MOUNT_JOURNAL_DATA; |
ac27a0ec DK |
1323 | goto datacheck; |
1324 | case Opt_data_ordered: | |
617ba13b | 1325 | data_opt = EXT4_MOUNT_ORDERED_DATA; |
ac27a0ec DK |
1326 | goto datacheck; |
1327 | case Opt_data_writeback: | |
617ba13b | 1328 | data_opt = EXT4_MOUNT_WRITEBACK_DATA; |
ac27a0ec DK |
1329 | datacheck: |
1330 | if (is_remount) { | |
617ba13b | 1331 | if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS) |
ac27a0ec | 1332 | != data_opt) { |
b31e1552 ES |
1333 | ext4_msg(sb, KERN_ERR, |
1334 | "Cannot change data mode on remount"); | |
ac27a0ec DK |
1335 | return 0; |
1336 | } | |
1337 | } else { | |
617ba13b | 1338 | sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS; |
ac27a0ec DK |
1339 | sbi->s_mount_opt |= data_opt; |
1340 | } | |
1341 | break; | |
5bf5683a HK |
1342 | case Opt_data_err_abort: |
1343 | set_opt(sbi->s_mount_opt, DATA_ERR_ABORT); | |
1344 | break; | |
1345 | case Opt_data_err_ignore: | |
1346 | clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT); | |
1347 | break; | |
f4033903 CW |
1348 | case Opt_mb_history_length: |
1349 | if (match_int(&args[0], &option)) | |
1350 | return 0; | |
1351 | if (option < 0) | |
1352 | return 0; | |
1353 | sbi->s_mb_history_max = option; | |
1354 | break; | |
ac27a0ec DK |
1355 | #ifdef CONFIG_QUOTA |
1356 | case Opt_usrjquota: | |
1357 | qtype = USRQUOTA; | |
1358 | goto set_qf_name; | |
1359 | case Opt_grpjquota: | |
1360 | qtype = GRPQUOTA; | |
1361 | set_qf_name: | |
17bd13b3 | 1362 | if (sb_any_quota_loaded(sb) && |
dfc5d03f | 1363 | !sbi->s_qf_names[qtype]) { |
b31e1552 ES |
1364 | ext4_msg(sb, KERN_ERR, |
1365 | "Cannot change journaled " | |
1366 | "quota options when quota turned on"); | |
ac27a0ec DK |
1367 | return 0; |
1368 | } | |
1369 | qname = match_strdup(&args[0]); | |
1370 | if (!qname) { | |
b31e1552 ES |
1371 | ext4_msg(sb, KERN_ERR, |
1372 | "Not enough memory for " | |
1373 | "storing quotafile name"); | |
ac27a0ec DK |
1374 | return 0; |
1375 | } | |
1376 | if (sbi->s_qf_names[qtype] && | |
1377 | strcmp(sbi->s_qf_names[qtype], qname)) { | |
b31e1552 ES |
1378 | ext4_msg(sb, KERN_ERR, |
1379 | "%s quota file already " | |
1380 | "specified", QTYPE2NAME(qtype)); | |
ac27a0ec DK |
1381 | kfree(qname); |
1382 | return 0; | |
1383 | } | |
1384 | sbi->s_qf_names[qtype] = qname; | |
1385 | if (strchr(sbi->s_qf_names[qtype], '/')) { | |
b31e1552 ES |
1386 | ext4_msg(sb, KERN_ERR, |
1387 | "quotafile must be on " | |
1388 | "filesystem root"); | |
ac27a0ec DK |
1389 | kfree(sbi->s_qf_names[qtype]); |
1390 | sbi->s_qf_names[qtype] = NULL; | |
1391 | return 0; | |
1392 | } | |
1393 | set_opt(sbi->s_mount_opt, QUOTA); | |
1394 | break; | |
1395 | case Opt_offusrjquota: | |
1396 | qtype = USRQUOTA; | |
1397 | goto clear_qf_name; | |
1398 | case Opt_offgrpjquota: | |
1399 | qtype = GRPQUOTA; | |
1400 | clear_qf_name: | |
17bd13b3 | 1401 | if (sb_any_quota_loaded(sb) && |
dfc5d03f | 1402 | sbi->s_qf_names[qtype]) { |
b31e1552 | 1403 | ext4_msg(sb, KERN_ERR, "Cannot change " |
2c8be6b2 | 1404 | "journaled quota options when " |
b31e1552 | 1405 | "quota turned on"); |
ac27a0ec DK |
1406 | return 0; |
1407 | } | |
1408 | /* | |
1409 | * The space will be released later when all options | |
1410 | * are confirmed to be correct | |
1411 | */ | |
1412 | sbi->s_qf_names[qtype] = NULL; | |
1413 | break; | |
1414 | case Opt_jqfmt_vfsold: | |
dfc5d03f JK |
1415 | qfmt = QFMT_VFS_OLD; |
1416 | goto set_qf_format; | |
ac27a0ec | 1417 | case Opt_jqfmt_vfsv0: |
dfc5d03f JK |
1418 | qfmt = QFMT_VFS_V0; |
1419 | set_qf_format: | |
17bd13b3 | 1420 | if (sb_any_quota_loaded(sb) && |
dfc5d03f | 1421 | sbi->s_jquota_fmt != qfmt) { |
b31e1552 | 1422 | ext4_msg(sb, KERN_ERR, "Cannot change " |
dfc5d03f | 1423 | "journaled quota options when " |
b31e1552 | 1424 | "quota turned on"); |
dfc5d03f JK |
1425 | return 0; |
1426 | } | |
1427 | sbi->s_jquota_fmt = qfmt; | |
ac27a0ec DK |
1428 | break; |
1429 | case Opt_quota: | |
1430 | case Opt_usrquota: | |
1431 | set_opt(sbi->s_mount_opt, QUOTA); | |
1432 | set_opt(sbi->s_mount_opt, USRQUOTA); | |
1433 | break; | |
1434 | case Opt_grpquota: | |
1435 | set_opt(sbi->s_mount_opt, QUOTA); | |
1436 | set_opt(sbi->s_mount_opt, GRPQUOTA); | |
1437 | break; | |
1438 | case Opt_noquota: | |
17bd13b3 | 1439 | if (sb_any_quota_loaded(sb)) { |
b31e1552 ES |
1440 | ext4_msg(sb, KERN_ERR, "Cannot change quota " |
1441 | "options when quota turned on"); | |
ac27a0ec DK |
1442 | return 0; |
1443 | } | |
1444 | clear_opt(sbi->s_mount_opt, QUOTA); | |
1445 | clear_opt(sbi->s_mount_opt, USRQUOTA); | |
1446 | clear_opt(sbi->s_mount_opt, GRPQUOTA); | |
1447 | break; | |
1448 | #else | |
1449 | case Opt_quota: | |
1450 | case Opt_usrquota: | |
1451 | case Opt_grpquota: | |
b31e1552 ES |
1452 | ext4_msg(sb, KERN_ERR, |
1453 | "quota options not supported"); | |
cd59e7b9 | 1454 | break; |
ac27a0ec DK |
1455 | case Opt_usrjquota: |
1456 | case Opt_grpjquota: | |
1457 | case Opt_offusrjquota: | |
1458 | case Opt_offgrpjquota: | |
1459 | case Opt_jqfmt_vfsold: | |
1460 | case Opt_jqfmt_vfsv0: | |
b31e1552 ES |
1461 | ext4_msg(sb, KERN_ERR, |
1462 | "journaled quota options not supported"); | |
ac27a0ec DK |
1463 | break; |
1464 | case Opt_noquota: | |
1465 | break; | |
1466 | #endif | |
1467 | case Opt_abort: | |
4ab2f15b | 1468 | sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; |
ac27a0ec | 1469 | break; |
06705bff TT |
1470 | case Opt_nobarrier: |
1471 | clear_opt(sbi->s_mount_opt, BARRIER); | |
1472 | break; | |
ac27a0ec | 1473 | case Opt_barrier: |
06705bff TT |
1474 | if (match_int(&args[0], &option)) { |
1475 | set_opt(sbi->s_mount_opt, BARRIER); | |
1476 | break; | |
1477 | } | |
ac27a0ec DK |
1478 | if (option) |
1479 | set_opt(sbi->s_mount_opt, BARRIER); | |
1480 | else | |
1481 | clear_opt(sbi->s_mount_opt, BARRIER); | |
1482 | break; | |
1483 | case Opt_ignore: | |
1484 | break; | |
1485 | case Opt_resize: | |
1486 | if (!is_remount) { | |
b31e1552 ES |
1487 | ext4_msg(sb, KERN_ERR, |
1488 | "resize option only available " | |
1489 | "for remount"); | |
ac27a0ec DK |
1490 | return 0; |
1491 | } | |
1492 | if (match_int(&args[0], &option) != 0) | |
1493 | return 0; | |
1494 | *n_blocks_count = option; | |
1495 | break; | |
1496 | case Opt_nobh: | |
1497 | set_opt(sbi->s_mount_opt, NOBH); | |
1498 | break; | |
1499 | case Opt_bh: | |
1500 | clear_opt(sbi->s_mount_opt, NOBH); | |
1501 | break; | |
25ec56b5 JNC |
1502 | case Opt_i_version: |
1503 | set_opt(sbi->s_mount_opt, I_VERSION); | |
1504 | sb->s_flags |= MS_I_VERSION; | |
1505 | break; | |
dd919b98 AK |
1506 | case Opt_nodelalloc: |
1507 | clear_opt(sbi->s_mount_opt, DELALLOC); | |
1508 | break; | |
c9de560d AT |
1509 | case Opt_stripe: |
1510 | if (match_int(&args[0], &option)) | |
1511 | return 0; | |
1512 | if (option < 0) | |
1513 | return 0; | |
1514 | sbi->s_stripe = option; | |
1515 | break; | |
64769240 AT |
1516 | case Opt_delalloc: |
1517 | set_opt(sbi->s_mount_opt, DELALLOC); | |
1518 | break; | |
6fd058f7 TT |
1519 | case Opt_block_validity: |
1520 | set_opt(sbi->s_mount_opt, BLOCK_VALIDITY); | |
1521 | break; | |
1522 | case Opt_noblock_validity: | |
1523 | clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY); | |
1524 | break; | |
240799cd TT |
1525 | case Opt_inode_readahead_blks: |
1526 | if (match_int(&args[0], &option)) | |
1527 | return 0; | |
1528 | if (option < 0 || option > (1 << 30)) | |
1529 | return 0; | |
f7c43950 | 1530 | if (!is_power_of_2(option)) { |
b31e1552 ES |
1531 | ext4_msg(sb, KERN_ERR, |
1532 | "EXT4-fs: inode_readahead_blks" | |
1533 | " must be a power of 2"); | |
3197ebdb TT |
1534 | return 0; |
1535 | } | |
240799cd TT |
1536 | sbi->s_inode_readahead_blks = option; |
1537 | break; | |
b3881f74 TT |
1538 | case Opt_journal_ioprio: |
1539 | if (match_int(&args[0], &option)) | |
1540 | return 0; | |
1541 | if (option < 0 || option > 7) | |
1542 | break; | |
1543 | *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, | |
1544 | option); | |
1545 | break; | |
06705bff TT |
1546 | case Opt_noauto_da_alloc: |
1547 | set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC); | |
1548 | break; | |
afd4672d | 1549 | case Opt_auto_da_alloc: |
06705bff TT |
1550 | if (match_int(&args[0], &option)) { |
1551 | clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC); | |
1552 | break; | |
1553 | } | |
afd4672d TT |
1554 | if (option) |
1555 | clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC); | |
1556 | else | |
1557 | set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC); | |
1558 | break; | |
ac27a0ec | 1559 | default: |
b31e1552 ES |
1560 | ext4_msg(sb, KERN_ERR, |
1561 | "Unrecognized mount option \"%s\" " | |
1562 | "or missing value", p); | |
ac27a0ec DK |
1563 | return 0; |
1564 | } | |
1565 | } | |
1566 | #ifdef CONFIG_QUOTA | |
1567 | if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { | |
617ba13b | 1568 | if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) && |
ac27a0ec DK |
1569 | sbi->s_qf_names[USRQUOTA]) |
1570 | clear_opt(sbi->s_mount_opt, USRQUOTA); | |
1571 | ||
617ba13b | 1572 | if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) && |
ac27a0ec DK |
1573 | sbi->s_qf_names[GRPQUOTA]) |
1574 | clear_opt(sbi->s_mount_opt, GRPQUOTA); | |
1575 | ||
1576 | if ((sbi->s_qf_names[USRQUOTA] && | |
617ba13b | 1577 | (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) || |
ac27a0ec | 1578 | (sbi->s_qf_names[GRPQUOTA] && |
617ba13b | 1579 | (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) { |
b31e1552 ES |
1580 | ext4_msg(sb, KERN_ERR, "old and new quota " |
1581 | "format mixing"); | |
ac27a0ec DK |
1582 | return 0; |
1583 | } | |
1584 | ||
1585 | if (!sbi->s_jquota_fmt) { | |
b31e1552 ES |
1586 | ext4_msg(sb, KERN_ERR, "journaled quota format " |
1587 | "not specified"); | |
ac27a0ec DK |
1588 | return 0; |
1589 | } | |
1590 | } else { | |
1591 | if (sbi->s_jquota_fmt) { | |
b31e1552 | 1592 | ext4_msg(sb, KERN_ERR, "journaled quota format " |
2c8be6b2 | 1593 | "specified with no journaling " |
b31e1552 | 1594 | "enabled"); |
ac27a0ec DK |
1595 | return 0; |
1596 | } | |
1597 | } | |
1598 | #endif | |
1599 | return 1; | |
1600 | } | |
1601 | ||
617ba13b | 1602 | static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, |
ac27a0ec DK |
1603 | int read_only) |
1604 | { | |
617ba13b | 1605 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec DK |
1606 | int res = 0; |
1607 | ||
617ba13b | 1608 | if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { |
b31e1552 ES |
1609 | ext4_msg(sb, KERN_ERR, "revision level too high, " |
1610 | "forcing read-only mode"); | |
ac27a0ec DK |
1611 | res = MS_RDONLY; |
1612 | } | |
1613 | if (read_only) | |
1614 | return res; | |
617ba13b | 1615 | if (!(sbi->s_mount_state & EXT4_VALID_FS)) |
b31e1552 ES |
1616 | ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " |
1617 | "running e2fsck is recommended"); | |
617ba13b | 1618 | else if ((sbi->s_mount_state & EXT4_ERROR_FS)) |
b31e1552 ES |
1619 | ext4_msg(sb, KERN_WARNING, |
1620 | "warning: mounting fs with errors, " | |
1621 | "running e2fsck is recommended"); | |
ac27a0ec DK |
1622 | else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 && |
1623 | le16_to_cpu(es->s_mnt_count) >= | |
1624 | (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) | |
b31e1552 ES |
1625 | ext4_msg(sb, KERN_WARNING, |
1626 | "warning: maximal mount count reached, " | |
1627 | "running e2fsck is recommended"); | |
ac27a0ec DK |
1628 | else if (le32_to_cpu(es->s_checkinterval) && |
1629 | (le32_to_cpu(es->s_lastcheck) + | |
1630 | le32_to_cpu(es->s_checkinterval) <= get_seconds())) | |
b31e1552 ES |
1631 | ext4_msg(sb, KERN_WARNING, |
1632 | "warning: checktime reached, " | |
1633 | "running e2fsck is recommended"); | |
0b8e58a1 | 1634 | if (!sbi->s_journal) |
0390131b | 1635 | es->s_state &= cpu_to_le16(~EXT4_VALID_FS); |
ac27a0ec | 1636 | if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) |
617ba13b | 1637 | es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); |
e8546d06 | 1638 | le16_add_cpu(&es->s_mnt_count, 1); |
ac27a0ec | 1639 | es->s_mtime = cpu_to_le32(get_seconds()); |
617ba13b | 1640 | ext4_update_dynamic_rev(sb); |
0390131b FM |
1641 | if (sbi->s_journal) |
1642 | EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
ac27a0ec | 1643 | |
e2d67052 | 1644 | ext4_commit_super(sb, 1); |
ac27a0ec | 1645 | if (test_opt(sb, DEBUG)) |
a9df9a49 | 1646 | printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " |
7f4520cc | 1647 | "bpg=%lu, ipg=%lu, mo=%04x]\n", |
ac27a0ec DK |
1648 | sb->s_blocksize, |
1649 | sbi->s_groups_count, | |
617ba13b MC |
1650 | EXT4_BLOCKS_PER_GROUP(sb), |
1651 | EXT4_INODES_PER_GROUP(sb), | |
ac27a0ec DK |
1652 | sbi->s_mount_opt); |
1653 | ||
0390131b | 1654 | if (EXT4_SB(sb)->s_journal) { |
b31e1552 ES |
1655 | ext4_msg(sb, KERN_INFO, "%s journal on %s", |
1656 | EXT4_SB(sb)->s_journal->j_inode ? "internal" : | |
0390131b FM |
1657 | "external", EXT4_SB(sb)->s_journal->j_devname); |
1658 | } else { | |
b31e1552 | 1659 | ext4_msg(sb, KERN_INFO, "no journal"); |
0390131b | 1660 | } |
ac27a0ec DK |
1661 | return res; |
1662 | } | |
1663 | ||
772cb7c8 JS |
1664 | static int ext4_fill_flex_info(struct super_block *sb) |
1665 | { | |
1666 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1667 | struct ext4_group_desc *gdp = NULL; | |
772cb7c8 JS |
1668 | ext4_group_t flex_group_count; |
1669 | ext4_group_t flex_group; | |
1670 | int groups_per_flex = 0; | |
c5ca7c76 | 1671 | size_t size; |
772cb7c8 JS |
1672 | int i; |
1673 | ||
1674 | if (!sbi->s_es->s_log_groups_per_flex) { | |
1675 | sbi->s_log_groups_per_flex = 0; | |
1676 | return 1; | |
1677 | } | |
1678 | ||
1679 | sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; | |
1680 | groups_per_flex = 1 << sbi->s_log_groups_per_flex; | |
1681 | ||
c62a11fd FB |
1682 | /* We allocate both existing and potentially added groups */ |
1683 | flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) + | |
d94e99a6 AK |
1684 | ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) << |
1685 | EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex; | |
c5ca7c76 TT |
1686 | size = flex_group_count * sizeof(struct flex_groups); |
1687 | sbi->s_flex_groups = kzalloc(size, GFP_KERNEL); | |
1688 | if (sbi->s_flex_groups == NULL) { | |
1689 | sbi->s_flex_groups = vmalloc(size); | |
1690 | if (sbi->s_flex_groups) | |
1691 | memset(sbi->s_flex_groups, 0, size); | |
1692 | } | |
772cb7c8 | 1693 | if (sbi->s_flex_groups == NULL) { |
b31e1552 ES |
1694 | ext4_msg(sb, KERN_ERR, "not enough memory for " |
1695 | "%u flex groups", flex_group_count); | |
772cb7c8 JS |
1696 | goto failed; |
1697 | } | |
772cb7c8 | 1698 | |
772cb7c8 | 1699 | for (i = 0; i < sbi->s_groups_count; i++) { |
88b6edd1 | 1700 | gdp = ext4_get_group_desc(sb, i, NULL); |
772cb7c8 JS |
1701 | |
1702 | flex_group = ext4_flex_group(sbi, i); | |
7ad9bb65 TT |
1703 | atomic_add(ext4_free_inodes_count(sb, gdp), |
1704 | &sbi->s_flex_groups[flex_group].free_inodes); | |
1705 | atomic_add(ext4_free_blks_count(sb, gdp), | |
1706 | &sbi->s_flex_groups[flex_group].free_blocks); | |
1707 | atomic_add(ext4_used_dirs_count(sb, gdp), | |
1708 | &sbi->s_flex_groups[flex_group].used_dirs); | |
772cb7c8 JS |
1709 | } |
1710 | ||
1711 | return 1; | |
1712 | failed: | |
1713 | return 0; | |
1714 | } | |
1715 | ||
717d50e4 AD |
1716 | __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group, |
1717 | struct ext4_group_desc *gdp) | |
1718 | { | |
1719 | __u16 crc = 0; | |
1720 | ||
1721 | if (sbi->s_es->s_feature_ro_compat & | |
1722 | cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) { | |
1723 | int offset = offsetof(struct ext4_group_desc, bg_checksum); | |
1724 | __le32 le_group = cpu_to_le32(block_group); | |
1725 | ||
1726 | crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); | |
1727 | crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); | |
1728 | crc = crc16(crc, (__u8 *)gdp, offset); | |
1729 | offset += sizeof(gdp->bg_checksum); /* skip checksum */ | |
1730 | /* for checksum of struct ext4_group_desc do the rest...*/ | |
1731 | if ((sbi->s_es->s_feature_incompat & | |
1732 | cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) && | |
1733 | offset < le16_to_cpu(sbi->s_es->s_desc_size)) | |
1734 | crc = crc16(crc, (__u8 *)gdp + offset, | |
1735 | le16_to_cpu(sbi->s_es->s_desc_size) - | |
1736 | offset); | |
1737 | } | |
1738 | ||
1739 | return cpu_to_le16(crc); | |
1740 | } | |
1741 | ||
1742 | int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group, | |
1743 | struct ext4_group_desc *gdp) | |
1744 | { | |
1745 | if ((sbi->s_es->s_feature_ro_compat & | |
1746 | cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) && | |
1747 | (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp))) | |
1748 | return 0; | |
1749 | ||
1750 | return 1; | |
1751 | } | |
1752 | ||
ac27a0ec | 1753 | /* Called at mount-time, super-block is locked */ |
197cd65a | 1754 | static int ext4_check_descriptors(struct super_block *sb) |
ac27a0ec | 1755 | { |
617ba13b MC |
1756 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
1757 | ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); | |
1758 | ext4_fsblk_t last_block; | |
bd81d8ee LV |
1759 | ext4_fsblk_t block_bitmap; |
1760 | ext4_fsblk_t inode_bitmap; | |
1761 | ext4_fsblk_t inode_table; | |
ce421581 | 1762 | int flexbg_flag = 0; |
fd2d4291 | 1763 | ext4_group_t i; |
ac27a0ec | 1764 | |
ce421581 JS |
1765 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) |
1766 | flexbg_flag = 1; | |
1767 | ||
af5bc92d | 1768 | ext4_debug("Checking group descriptors"); |
ac27a0ec | 1769 | |
197cd65a AM |
1770 | for (i = 0; i < sbi->s_groups_count; i++) { |
1771 | struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); | |
1772 | ||
ce421581 | 1773 | if (i == sbi->s_groups_count - 1 || flexbg_flag) |
bd81d8ee | 1774 | last_block = ext4_blocks_count(sbi->s_es) - 1; |
ac27a0ec DK |
1775 | else |
1776 | last_block = first_block + | |
617ba13b | 1777 | (EXT4_BLOCKS_PER_GROUP(sb) - 1); |
ac27a0ec | 1778 | |
8fadc143 | 1779 | block_bitmap = ext4_block_bitmap(sb, gdp); |
2b2d6d01 | 1780 | if (block_bitmap < first_block || block_bitmap > last_block) { |
b31e1552 | 1781 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 1782 | "Block bitmap for group %u not in group " |
b31e1552 | 1783 | "(block %llu)!", i, block_bitmap); |
ac27a0ec DK |
1784 | return 0; |
1785 | } | |
8fadc143 | 1786 | inode_bitmap = ext4_inode_bitmap(sb, gdp); |
2b2d6d01 | 1787 | if (inode_bitmap < first_block || inode_bitmap > last_block) { |
b31e1552 | 1788 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 1789 | "Inode bitmap for group %u not in group " |
b31e1552 | 1790 | "(block %llu)!", i, inode_bitmap); |
ac27a0ec DK |
1791 | return 0; |
1792 | } | |
8fadc143 | 1793 | inode_table = ext4_inode_table(sb, gdp); |
bd81d8ee | 1794 | if (inode_table < first_block || |
2b2d6d01 | 1795 | inode_table + sbi->s_itb_per_group - 1 > last_block) { |
b31e1552 | 1796 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 1797 | "Inode table for group %u not in group " |
b31e1552 | 1798 | "(block %llu)!", i, inode_table); |
ac27a0ec DK |
1799 | return 0; |
1800 | } | |
955ce5f5 | 1801 | ext4_lock_group(sb, i); |
717d50e4 | 1802 | if (!ext4_group_desc_csum_verify(sbi, i, gdp)) { |
b31e1552 ES |
1803 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
1804 | "Checksum for group %u failed (%u!=%u)", | |
1805 | i, le16_to_cpu(ext4_group_desc_csum(sbi, i, | |
1806 | gdp)), le16_to_cpu(gdp->bg_checksum)); | |
7ee1ec4c | 1807 | if (!(sb->s_flags & MS_RDONLY)) { |
955ce5f5 | 1808 | ext4_unlock_group(sb, i); |
8a266467 | 1809 | return 0; |
7ee1ec4c | 1810 | } |
717d50e4 | 1811 | } |
955ce5f5 | 1812 | ext4_unlock_group(sb, i); |
ce421581 JS |
1813 | if (!flexbg_flag) |
1814 | first_block += EXT4_BLOCKS_PER_GROUP(sb); | |
ac27a0ec DK |
1815 | } |
1816 | ||
bd81d8ee | 1817 | ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb)); |
0b8e58a1 | 1818 | sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb)); |
ac27a0ec DK |
1819 | return 1; |
1820 | } | |
1821 | ||
617ba13b | 1822 | /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at |
ac27a0ec DK |
1823 | * the superblock) which were deleted from all directories, but held open by |
1824 | * a process at the time of a crash. We walk the list and try to delete these | |
1825 | * inodes at recovery time (only with a read-write filesystem). | |
1826 | * | |
1827 | * In order to keep the orphan inode chain consistent during traversal (in | |
1828 | * case of crash during recovery), we link each inode into the superblock | |
1829 | * orphan list_head and handle it the same way as an inode deletion during | |
1830 | * normal operation (which journals the operations for us). | |
1831 | * | |
1832 | * We only do an iget() and an iput() on each inode, which is very safe if we | |
1833 | * accidentally point at an in-use or already deleted inode. The worst that | |
1834 | * can happen in this case is that we get a "bit already cleared" message from | |
617ba13b | 1835 | * ext4_free_inode(). The only reason we would point at a wrong inode is if |
ac27a0ec DK |
1836 | * e2fsck was run on this filesystem, and it must have already done the orphan |
1837 | * inode cleanup for us, so we can safely abort without any further action. | |
1838 | */ | |
2b2d6d01 TT |
1839 | static void ext4_orphan_cleanup(struct super_block *sb, |
1840 | struct ext4_super_block *es) | |
ac27a0ec DK |
1841 | { |
1842 | unsigned int s_flags = sb->s_flags; | |
1843 | int nr_orphans = 0, nr_truncates = 0; | |
1844 | #ifdef CONFIG_QUOTA | |
1845 | int i; | |
1846 | #endif | |
1847 | if (!es->s_last_orphan) { | |
1848 | jbd_debug(4, "no orphan inodes to clean up\n"); | |
1849 | return; | |
1850 | } | |
1851 | ||
a8f48a95 | 1852 | if (bdev_read_only(sb->s_bdev)) { |
b31e1552 ES |
1853 | ext4_msg(sb, KERN_ERR, "write access " |
1854 | "unavailable, skipping orphan cleanup"); | |
a8f48a95 ES |
1855 | return; |
1856 | } | |
1857 | ||
617ba13b | 1858 | if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { |
ac27a0ec DK |
1859 | if (es->s_last_orphan) |
1860 | jbd_debug(1, "Errors on filesystem, " | |
1861 | "clearing orphan list.\n"); | |
1862 | es->s_last_orphan = 0; | |
1863 | jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); | |
1864 | return; | |
1865 | } | |
1866 | ||
1867 | if (s_flags & MS_RDONLY) { | |
b31e1552 | 1868 | ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); |
ac27a0ec DK |
1869 | sb->s_flags &= ~MS_RDONLY; |
1870 | } | |
1871 | #ifdef CONFIG_QUOTA | |
1872 | /* Needed for iput() to work correctly and not trash data */ | |
1873 | sb->s_flags |= MS_ACTIVE; | |
1874 | /* Turn on quotas so that they are updated correctly */ | |
1875 | for (i = 0; i < MAXQUOTAS; i++) { | |
617ba13b MC |
1876 | if (EXT4_SB(sb)->s_qf_names[i]) { |
1877 | int ret = ext4_quota_on_mount(sb, i); | |
ac27a0ec | 1878 | if (ret < 0) |
b31e1552 ES |
1879 | ext4_msg(sb, KERN_ERR, |
1880 | "Cannot turn on journaled " | |
1881 | "quota: error %d", ret); | |
ac27a0ec DK |
1882 | } |
1883 | } | |
1884 | #endif | |
1885 | ||
1886 | while (es->s_last_orphan) { | |
1887 | struct inode *inode; | |
1888 | ||
97bd42b9 JB |
1889 | inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); |
1890 | if (IS_ERR(inode)) { | |
ac27a0ec DK |
1891 | es->s_last_orphan = 0; |
1892 | break; | |
1893 | } | |
1894 | ||
617ba13b | 1895 | list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); |
a269eb18 | 1896 | vfs_dq_init(inode); |
ac27a0ec | 1897 | if (inode->i_nlink) { |
b31e1552 ES |
1898 | ext4_msg(sb, KERN_DEBUG, |
1899 | "%s: truncating inode %lu to %lld bytes", | |
46e665e9 | 1900 | __func__, inode->i_ino, inode->i_size); |
e5f8eab8 | 1901 | jbd_debug(2, "truncating inode %lu to %lld bytes\n", |
ac27a0ec | 1902 | inode->i_ino, inode->i_size); |
617ba13b | 1903 | ext4_truncate(inode); |
ac27a0ec DK |
1904 | nr_truncates++; |
1905 | } else { | |
b31e1552 ES |
1906 | ext4_msg(sb, KERN_DEBUG, |
1907 | "%s: deleting unreferenced inode %lu", | |
46e665e9 | 1908 | __func__, inode->i_ino); |
ac27a0ec DK |
1909 | jbd_debug(2, "deleting unreferenced inode %lu\n", |
1910 | inode->i_ino); | |
1911 | nr_orphans++; | |
1912 | } | |
1913 | iput(inode); /* The delete magic happens here! */ | |
1914 | } | |
1915 | ||
2b2d6d01 | 1916 | #define PLURAL(x) (x), ((x) == 1) ? "" : "s" |
ac27a0ec DK |
1917 | |
1918 | if (nr_orphans) | |
b31e1552 ES |
1919 | ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", |
1920 | PLURAL(nr_orphans)); | |
ac27a0ec | 1921 | if (nr_truncates) |
b31e1552 ES |
1922 | ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", |
1923 | PLURAL(nr_truncates)); | |
ac27a0ec DK |
1924 | #ifdef CONFIG_QUOTA |
1925 | /* Turn quotas off */ | |
1926 | for (i = 0; i < MAXQUOTAS; i++) { | |
1927 | if (sb_dqopt(sb)->files[i]) | |
6f28e087 | 1928 | vfs_quota_off(sb, i, 0); |
ac27a0ec DK |
1929 | } |
1930 | #endif | |
1931 | sb->s_flags = s_flags; /* Restore MS_RDONLY status */ | |
1932 | } | |
0b8e58a1 | 1933 | |
cd2291a4 ES |
1934 | /* |
1935 | * Maximal extent format file size. | |
1936 | * Resulting logical blkno at s_maxbytes must fit in our on-disk | |
1937 | * extent format containers, within a sector_t, and within i_blocks | |
1938 | * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, | |
1939 | * so that won't be a limiting factor. | |
1940 | * | |
1941 | * Note, this does *not* consider any metadata overhead for vfs i_blocks. | |
1942 | */ | |
f287a1a5 | 1943 | static loff_t ext4_max_size(int blkbits, int has_huge_files) |
cd2291a4 ES |
1944 | { |
1945 | loff_t res; | |
1946 | loff_t upper_limit = MAX_LFS_FILESIZE; | |
1947 | ||
1948 | /* small i_blocks in vfs inode? */ | |
f287a1a5 | 1949 | if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { |
cd2291a4 | 1950 | /* |
90c699a9 | 1951 | * CONFIG_LBDAF is not enabled implies the inode |
cd2291a4 ES |
1952 | * i_block represent total blocks in 512 bytes |
1953 | * 32 == size of vfs inode i_blocks * 8 | |
1954 | */ | |
1955 | upper_limit = (1LL << 32) - 1; | |
1956 | ||
1957 | /* total blocks in file system block size */ | |
1958 | upper_limit >>= (blkbits - 9); | |
1959 | upper_limit <<= blkbits; | |
1960 | } | |
1961 | ||
1962 | /* 32-bit extent-start container, ee_block */ | |
1963 | res = 1LL << 32; | |
1964 | res <<= blkbits; | |
1965 | res -= 1; | |
1966 | ||
1967 | /* Sanity check against vm- & vfs- imposed limits */ | |
1968 | if (res > upper_limit) | |
1969 | res = upper_limit; | |
1970 | ||
1971 | return res; | |
1972 | } | |
ac27a0ec | 1973 | |
ac27a0ec | 1974 | /* |
cd2291a4 | 1975 | * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect |
0fc1b451 AK |
1976 | * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. |
1977 | * We need to be 1 filesystem block less than the 2^48 sector limit. | |
ac27a0ec | 1978 | */ |
f287a1a5 | 1979 | static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) |
ac27a0ec | 1980 | { |
617ba13b | 1981 | loff_t res = EXT4_NDIR_BLOCKS; |
0fc1b451 AK |
1982 | int meta_blocks; |
1983 | loff_t upper_limit; | |
0b8e58a1 AD |
1984 | /* This is calculated to be the largest file size for a dense, block |
1985 | * mapped file such that the file's total number of 512-byte sectors, | |
1986 | * including data and all indirect blocks, does not exceed (2^48 - 1). | |
1987 | * | |
1988 | * __u32 i_blocks_lo and _u16 i_blocks_high represent the total | |
1989 | * number of 512-byte sectors of the file. | |
0fc1b451 AK |
1990 | */ |
1991 | ||
f287a1a5 | 1992 | if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { |
0fc1b451 | 1993 | /* |
90c699a9 | 1994 | * !has_huge_files or CONFIG_LBDAF not enabled implies that |
0b8e58a1 AD |
1995 | * the inode i_block field represents total file blocks in |
1996 | * 2^32 512-byte sectors == size of vfs inode i_blocks * 8 | |
0fc1b451 AK |
1997 | */ |
1998 | upper_limit = (1LL << 32) - 1; | |
1999 | ||
2000 | /* total blocks in file system block size */ | |
2001 | upper_limit >>= (bits - 9); | |
2002 | ||
2003 | } else { | |
8180a562 AK |
2004 | /* |
2005 | * We use 48 bit ext4_inode i_blocks | |
2006 | * With EXT4_HUGE_FILE_FL set the i_blocks | |
2007 | * represent total number of blocks in | |
2008 | * file system block size | |
2009 | */ | |
0fc1b451 AK |
2010 | upper_limit = (1LL << 48) - 1; |
2011 | ||
0fc1b451 AK |
2012 | } |
2013 | ||
2014 | /* indirect blocks */ | |
2015 | meta_blocks = 1; | |
2016 | /* double indirect blocks */ | |
2017 | meta_blocks += 1 + (1LL << (bits-2)); | |
2018 | /* tripple indirect blocks */ | |
2019 | meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); | |
2020 | ||
2021 | upper_limit -= meta_blocks; | |
2022 | upper_limit <<= bits; | |
ac27a0ec DK |
2023 | |
2024 | res += 1LL << (bits-2); | |
2025 | res += 1LL << (2*(bits-2)); | |
2026 | res += 1LL << (3*(bits-2)); | |
2027 | res <<= bits; | |
2028 | if (res > upper_limit) | |
2029 | res = upper_limit; | |
0fc1b451 AK |
2030 | |
2031 | if (res > MAX_LFS_FILESIZE) | |
2032 | res = MAX_LFS_FILESIZE; | |
2033 | ||
ac27a0ec DK |
2034 | return res; |
2035 | } | |
2036 | ||
617ba13b | 2037 | static ext4_fsblk_t descriptor_loc(struct super_block *sb, |
0b8e58a1 | 2038 | ext4_fsblk_t logical_sb_block, int nr) |
ac27a0ec | 2039 | { |
617ba13b | 2040 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
fd2d4291 | 2041 | ext4_group_t bg, first_meta_bg; |
ac27a0ec DK |
2042 | int has_super = 0; |
2043 | ||
2044 | first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); | |
2045 | ||
617ba13b | 2046 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) || |
ac27a0ec | 2047 | nr < first_meta_bg) |
70bbb3e0 | 2048 | return logical_sb_block + nr + 1; |
ac27a0ec | 2049 | bg = sbi->s_desc_per_block * nr; |
617ba13b | 2050 | if (ext4_bg_has_super(sb, bg)) |
ac27a0ec | 2051 | has_super = 1; |
0b8e58a1 | 2052 | |
617ba13b | 2053 | return (has_super + ext4_group_first_block_no(sb, bg)); |
ac27a0ec DK |
2054 | } |
2055 | ||
c9de560d AT |
2056 | /** |
2057 | * ext4_get_stripe_size: Get the stripe size. | |
2058 | * @sbi: In memory super block info | |
2059 | * | |
2060 | * If we have specified it via mount option, then | |
2061 | * use the mount option value. If the value specified at mount time is | |
2062 | * greater than the blocks per group use the super block value. | |
2063 | * If the super block value is greater than blocks per group return 0. | |
2064 | * Allocator needs it be less than blocks per group. | |
2065 | * | |
2066 | */ | |
2067 | static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) | |
2068 | { | |
2069 | unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); | |
2070 | unsigned long stripe_width = | |
2071 | le32_to_cpu(sbi->s_es->s_raid_stripe_width); | |
2072 | ||
2073 | if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) | |
2074 | return sbi->s_stripe; | |
2075 | ||
2076 | if (stripe_width <= sbi->s_blocks_per_group) | |
2077 | return stripe_width; | |
2078 | ||
2079 | if (stride <= sbi->s_blocks_per_group) | |
2080 | return stride; | |
2081 | ||
2082 | return 0; | |
2083 | } | |
ac27a0ec | 2084 | |
3197ebdb TT |
2085 | /* sysfs supprt */ |
2086 | ||
2087 | struct ext4_attr { | |
2088 | struct attribute attr; | |
2089 | ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *); | |
2090 | ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *, | |
2091 | const char *, size_t); | |
2092 | int offset; | |
2093 | }; | |
2094 | ||
2095 | static int parse_strtoul(const char *buf, | |
2096 | unsigned long max, unsigned long *value) | |
2097 | { | |
2098 | char *endp; | |
2099 | ||
2100 | while (*buf && isspace(*buf)) | |
2101 | buf++; | |
2102 | *value = simple_strtoul(buf, &endp, 0); | |
2103 | while (*endp && isspace(*endp)) | |
2104 | endp++; | |
2105 | if (*endp || *value > max) | |
2106 | return -EINVAL; | |
2107 | ||
2108 | return 0; | |
2109 | } | |
2110 | ||
2111 | static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a, | |
2112 | struct ext4_sb_info *sbi, | |
2113 | char *buf) | |
2114 | { | |
2115 | return snprintf(buf, PAGE_SIZE, "%llu\n", | |
2116 | (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter)); | |
2117 | } | |
2118 | ||
2119 | static ssize_t session_write_kbytes_show(struct ext4_attr *a, | |
2120 | struct ext4_sb_info *sbi, char *buf) | |
2121 | { | |
2122 | struct super_block *sb = sbi->s_buddy_cache->i_sb; | |
2123 | ||
2124 | return snprintf(buf, PAGE_SIZE, "%lu\n", | |
2125 | (part_stat_read(sb->s_bdev->bd_part, sectors[1]) - | |
2126 | sbi->s_sectors_written_start) >> 1); | |
2127 | } | |
2128 | ||
2129 | static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a, | |
2130 | struct ext4_sb_info *sbi, char *buf) | |
2131 | { | |
2132 | struct super_block *sb = sbi->s_buddy_cache->i_sb; | |
2133 | ||
2134 | return snprintf(buf, PAGE_SIZE, "%llu\n", | |
2135 | sbi->s_kbytes_written + | |
2136 | ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - | |
2137 | EXT4_SB(sb)->s_sectors_written_start) >> 1)); | |
2138 | } | |
2139 | ||
2140 | static ssize_t inode_readahead_blks_store(struct ext4_attr *a, | |
2141 | struct ext4_sb_info *sbi, | |
2142 | const char *buf, size_t count) | |
2143 | { | |
2144 | unsigned long t; | |
2145 | ||
2146 | if (parse_strtoul(buf, 0x40000000, &t)) | |
2147 | return -EINVAL; | |
2148 | ||
f7c43950 | 2149 | if (!is_power_of_2(t)) |
3197ebdb TT |
2150 | return -EINVAL; |
2151 | ||
2152 | sbi->s_inode_readahead_blks = t; | |
2153 | return count; | |
2154 | } | |
2155 | ||
2156 | static ssize_t sbi_ui_show(struct ext4_attr *a, | |
0b8e58a1 | 2157 | struct ext4_sb_info *sbi, char *buf) |
3197ebdb TT |
2158 | { |
2159 | unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset); | |
2160 | ||
2161 | return snprintf(buf, PAGE_SIZE, "%u\n", *ui); | |
2162 | } | |
2163 | ||
2164 | static ssize_t sbi_ui_store(struct ext4_attr *a, | |
2165 | struct ext4_sb_info *sbi, | |
2166 | const char *buf, size_t count) | |
2167 | { | |
2168 | unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset); | |
2169 | unsigned long t; | |
2170 | ||
2171 | if (parse_strtoul(buf, 0xffffffff, &t)) | |
2172 | return -EINVAL; | |
2173 | *ui = t; | |
2174 | return count; | |
2175 | } | |
2176 | ||
2177 | #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \ | |
2178 | static struct ext4_attr ext4_attr_##_name = { \ | |
2179 | .attr = {.name = __stringify(_name), .mode = _mode }, \ | |
2180 | .show = _show, \ | |
2181 | .store = _store, \ | |
2182 | .offset = offsetof(struct ext4_sb_info, _elname), \ | |
2183 | } | |
2184 | #define EXT4_ATTR(name, mode, show, store) \ | |
2185 | static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store) | |
2186 | ||
2187 | #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL) | |
2188 | #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store) | |
2189 | #define EXT4_RW_ATTR_SBI_UI(name, elname) \ | |
2190 | EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname) | |
2191 | #define ATTR_LIST(name) &ext4_attr_##name.attr | |
2192 | ||
2193 | EXT4_RO_ATTR(delayed_allocation_blocks); | |
2194 | EXT4_RO_ATTR(session_write_kbytes); | |
2195 | EXT4_RO_ATTR(lifetime_write_kbytes); | |
2196 | EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show, | |
2197 | inode_readahead_blks_store, s_inode_readahead_blks); | |
11013911 | 2198 | EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal); |
3197ebdb TT |
2199 | EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats); |
2200 | EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan); | |
2201 | EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan); | |
2202 | EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs); | |
2203 | EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request); | |
2204 | EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc); | |
55138e0b | 2205 | EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump); |
3197ebdb TT |
2206 | |
2207 | static struct attribute *ext4_attrs[] = { | |
2208 | ATTR_LIST(delayed_allocation_blocks), | |
2209 | ATTR_LIST(session_write_kbytes), | |
2210 | ATTR_LIST(lifetime_write_kbytes), | |
2211 | ATTR_LIST(inode_readahead_blks), | |
11013911 | 2212 | ATTR_LIST(inode_goal), |
3197ebdb TT |
2213 | ATTR_LIST(mb_stats), |
2214 | ATTR_LIST(mb_max_to_scan), | |
2215 | ATTR_LIST(mb_min_to_scan), | |
2216 | ATTR_LIST(mb_order2_req), | |
2217 | ATTR_LIST(mb_stream_req), | |
2218 | ATTR_LIST(mb_group_prealloc), | |
55138e0b | 2219 | ATTR_LIST(max_writeback_mb_bump), |
3197ebdb TT |
2220 | NULL, |
2221 | }; | |
2222 | ||
2223 | static ssize_t ext4_attr_show(struct kobject *kobj, | |
2224 | struct attribute *attr, char *buf) | |
2225 | { | |
2226 | struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, | |
2227 | s_kobj); | |
2228 | struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); | |
2229 | ||
2230 | return a->show ? a->show(a, sbi, buf) : 0; | |
2231 | } | |
2232 | ||
2233 | static ssize_t ext4_attr_store(struct kobject *kobj, | |
2234 | struct attribute *attr, | |
2235 | const char *buf, size_t len) | |
2236 | { | |
2237 | struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, | |
2238 | s_kobj); | |
2239 | struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); | |
2240 | ||
2241 | return a->store ? a->store(a, sbi, buf, len) : 0; | |
2242 | } | |
2243 | ||
2244 | static void ext4_sb_release(struct kobject *kobj) | |
2245 | { | |
2246 | struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, | |
2247 | s_kobj); | |
2248 | complete(&sbi->s_kobj_unregister); | |
2249 | } | |
2250 | ||
2251 | ||
2252 | static struct sysfs_ops ext4_attr_ops = { | |
2253 | .show = ext4_attr_show, | |
2254 | .store = ext4_attr_store, | |
2255 | }; | |
2256 | ||
2257 | static struct kobj_type ext4_ktype = { | |
2258 | .default_attrs = ext4_attrs, | |
2259 | .sysfs_ops = &ext4_attr_ops, | |
2260 | .release = ext4_sb_release, | |
2261 | }; | |
2262 | ||
a13fb1a4 ES |
2263 | /* |
2264 | * Check whether this filesystem can be mounted based on | |
2265 | * the features present and the RDONLY/RDWR mount requested. | |
2266 | * Returns 1 if this filesystem can be mounted as requested, | |
2267 | * 0 if it cannot be. | |
2268 | */ | |
2269 | static int ext4_feature_set_ok(struct super_block *sb, int readonly) | |
2270 | { | |
2271 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) { | |
2272 | ext4_msg(sb, KERN_ERR, | |
2273 | "Couldn't mount because of " | |
2274 | "unsupported optional features (%x)", | |
2275 | (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & | |
2276 | ~EXT4_FEATURE_INCOMPAT_SUPP)); | |
2277 | return 0; | |
2278 | } | |
2279 | ||
2280 | if (readonly) | |
2281 | return 1; | |
2282 | ||
2283 | /* Check that feature set is OK for a read-write mount */ | |
2284 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) { | |
2285 | ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " | |
2286 | "unsupported optional features (%x)", | |
2287 | (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & | |
2288 | ~EXT4_FEATURE_RO_COMPAT_SUPP)); | |
2289 | return 0; | |
2290 | } | |
2291 | /* | |
2292 | * Large file size enabled file system can only be mounted | |
2293 | * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF | |
2294 | */ | |
2295 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) { | |
2296 | if (sizeof(blkcnt_t) < sizeof(u64)) { | |
2297 | ext4_msg(sb, KERN_ERR, "Filesystem with huge files " | |
2298 | "cannot be mounted RDWR without " | |
2299 | "CONFIG_LBDAF"); | |
2300 | return 0; | |
2301 | } | |
2302 | } | |
2303 | return 1; | |
2304 | } | |
2305 | ||
2b2d6d01 | 2306 | static int ext4_fill_super(struct super_block *sb, void *data, int silent) |
7477827f AK |
2307 | __releases(kernel_lock) |
2308 | __acquires(kernel_lock) | |
ac27a0ec | 2309 | { |
2b2d6d01 | 2310 | struct buffer_head *bh; |
617ba13b MC |
2311 | struct ext4_super_block *es = NULL; |
2312 | struct ext4_sb_info *sbi; | |
2313 | ext4_fsblk_t block; | |
2314 | ext4_fsblk_t sb_block = get_sb_block(&data); | |
70bbb3e0 | 2315 | ext4_fsblk_t logical_sb_block; |
ac27a0ec | 2316 | unsigned long offset = 0; |
ac27a0ec DK |
2317 | unsigned long journal_devnum = 0; |
2318 | unsigned long def_mount_opts; | |
2319 | struct inode *root; | |
9f6200bb | 2320 | char *cp; |
0390131b | 2321 | const char *descr; |
1d1fe1ee | 2322 | int ret = -EINVAL; |
ac27a0ec | 2323 | int blocksize; |
4ec11028 TT |
2324 | unsigned int db_count; |
2325 | unsigned int i; | |
f287a1a5 | 2326 | int needs_recovery, has_huge_files; |
bd81d8ee | 2327 | __u64 blocks_count; |
833f4077 | 2328 | int err; |
b3881f74 | 2329 | unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; |
ac27a0ec DK |
2330 | |
2331 | sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); | |
2332 | if (!sbi) | |
2333 | return -ENOMEM; | |
705895b6 PE |
2334 | |
2335 | sbi->s_blockgroup_lock = | |
2336 | kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); | |
2337 | if (!sbi->s_blockgroup_lock) { | |
2338 | kfree(sbi); | |
2339 | return -ENOMEM; | |
2340 | } | |
ac27a0ec DK |
2341 | sb->s_fs_info = sbi; |
2342 | sbi->s_mount_opt = 0; | |
617ba13b MC |
2343 | sbi->s_resuid = EXT4_DEF_RESUID; |
2344 | sbi->s_resgid = EXT4_DEF_RESGID; | |
240799cd | 2345 | sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; |
d9c9bef1 | 2346 | sbi->s_sb_block = sb_block; |
afc32f7e TT |
2347 | sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part, |
2348 | sectors[1]); | |
ac27a0ec DK |
2349 | |
2350 | unlock_kernel(); | |
2351 | ||
9f6200bb TT |
2352 | /* Cleanup superblock name */ |
2353 | for (cp = sb->s_id; (cp = strchr(cp, '/'));) | |
2354 | *cp = '!'; | |
2355 | ||
617ba13b | 2356 | blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); |
ac27a0ec | 2357 | if (!blocksize) { |
b31e1552 | 2358 | ext4_msg(sb, KERN_ERR, "unable to set blocksize"); |
ac27a0ec DK |
2359 | goto out_fail; |
2360 | } | |
2361 | ||
2362 | /* | |
617ba13b | 2363 | * The ext4 superblock will not be buffer aligned for other than 1kB |
ac27a0ec DK |
2364 | * block sizes. We need to calculate the offset from buffer start. |
2365 | */ | |
617ba13b | 2366 | if (blocksize != EXT4_MIN_BLOCK_SIZE) { |
70bbb3e0 AM |
2367 | logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; |
2368 | offset = do_div(logical_sb_block, blocksize); | |
ac27a0ec | 2369 | } else { |
70bbb3e0 | 2370 | logical_sb_block = sb_block; |
ac27a0ec DK |
2371 | } |
2372 | ||
70bbb3e0 | 2373 | if (!(bh = sb_bread(sb, logical_sb_block))) { |
b31e1552 | 2374 | ext4_msg(sb, KERN_ERR, "unable to read superblock"); |
ac27a0ec DK |
2375 | goto out_fail; |
2376 | } | |
2377 | /* | |
2378 | * Note: s_es must be initialized as soon as possible because | |
617ba13b | 2379 | * some ext4 macro-instructions depend on its value |
ac27a0ec | 2380 | */ |
617ba13b | 2381 | es = (struct ext4_super_block *) (((char *)bh->b_data) + offset); |
ac27a0ec DK |
2382 | sbi->s_es = es; |
2383 | sb->s_magic = le16_to_cpu(es->s_magic); | |
617ba13b MC |
2384 | if (sb->s_magic != EXT4_SUPER_MAGIC) |
2385 | goto cantfind_ext4; | |
afc32f7e | 2386 | sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); |
ac27a0ec DK |
2387 | |
2388 | /* Set defaults before we parse the mount options */ | |
2389 | def_mount_opts = le32_to_cpu(es->s_default_mount_opts); | |
617ba13b | 2390 | if (def_mount_opts & EXT4_DEFM_DEBUG) |
ac27a0ec | 2391 | set_opt(sbi->s_mount_opt, DEBUG); |
617ba13b | 2392 | if (def_mount_opts & EXT4_DEFM_BSDGROUPS) |
ac27a0ec | 2393 | set_opt(sbi->s_mount_opt, GRPID); |
617ba13b | 2394 | if (def_mount_opts & EXT4_DEFM_UID16) |
ac27a0ec | 2395 | set_opt(sbi->s_mount_opt, NO_UID32); |
03010a33 | 2396 | #ifdef CONFIG_EXT4_FS_XATTR |
617ba13b | 2397 | if (def_mount_opts & EXT4_DEFM_XATTR_USER) |
ac27a0ec | 2398 | set_opt(sbi->s_mount_opt, XATTR_USER); |
2e7842b8 | 2399 | #endif |
03010a33 | 2400 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
617ba13b | 2401 | if (def_mount_opts & EXT4_DEFM_ACL) |
ac27a0ec | 2402 | set_opt(sbi->s_mount_opt, POSIX_ACL); |
2e7842b8 | 2403 | #endif |
617ba13b MC |
2404 | if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) |
2405 | sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA; | |
2406 | else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) | |
2407 | sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA; | |
2408 | else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) | |
2409 | sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA; | |
2410 | ||
2411 | if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) | |
ac27a0ec | 2412 | set_opt(sbi->s_mount_opt, ERRORS_PANIC); |
bb4f397a | 2413 | else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) |
ceea16bf | 2414 | set_opt(sbi->s_mount_opt, ERRORS_CONT); |
bb4f397a AK |
2415 | else |
2416 | set_opt(sbi->s_mount_opt, ERRORS_RO); | |
ac27a0ec DK |
2417 | |
2418 | sbi->s_resuid = le16_to_cpu(es->s_def_resuid); | |
2419 | sbi->s_resgid = le16_to_cpu(es->s_def_resgid); | |
30773840 TT |
2420 | sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; |
2421 | sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; | |
2422 | sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; | |
f4033903 | 2423 | sbi->s_mb_history_max = default_mb_history_length; |
ac27a0ec | 2424 | |
571640ca | 2425 | set_opt(sbi->s_mount_opt, BARRIER); |
ac27a0ec | 2426 | |
dd919b98 AK |
2427 | /* |
2428 | * enable delayed allocation by default | |
2429 | * Use -o nodelalloc to turn it off | |
2430 | */ | |
2431 | set_opt(sbi->s_mount_opt, DELALLOC); | |
2432 | ||
b3881f74 TT |
2433 | if (!parse_options((char *) data, sb, &journal_devnum, |
2434 | &journal_ioprio, NULL, 0)) | |
ac27a0ec DK |
2435 | goto failed_mount; |
2436 | ||
2437 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
617ba13b | 2438 | ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0); |
ac27a0ec | 2439 | |
617ba13b MC |
2440 | if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && |
2441 | (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) || | |
2442 | EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) || | |
2443 | EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U))) | |
b31e1552 ES |
2444 | ext4_msg(sb, KERN_WARNING, |
2445 | "feature flags set on rev 0 fs, " | |
2446 | "running e2fsck is recommended"); | |
469108ff | 2447 | |
ac27a0ec DK |
2448 | /* |
2449 | * Check feature flags regardless of the revision level, since we | |
2450 | * previously didn't change the revision level when setting the flags, | |
2451 | * so there is a chance incompat flags are set on a rev 0 filesystem. | |
2452 | */ | |
a13fb1a4 | 2453 | if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY))) |
ac27a0ec | 2454 | goto failed_mount; |
a13fb1a4 | 2455 | |
ac27a0ec DK |
2456 | blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); |
2457 | ||
617ba13b MC |
2458 | if (blocksize < EXT4_MIN_BLOCK_SIZE || |
2459 | blocksize > EXT4_MAX_BLOCK_SIZE) { | |
b31e1552 ES |
2460 | ext4_msg(sb, KERN_ERR, |
2461 | "Unsupported filesystem blocksize %d", blocksize); | |
ac27a0ec DK |
2462 | goto failed_mount; |
2463 | } | |
2464 | ||
ac27a0ec | 2465 | if (sb->s_blocksize != blocksize) { |
ce40733c AK |
2466 | /* Validate the filesystem blocksize */ |
2467 | if (!sb_set_blocksize(sb, blocksize)) { | |
b31e1552 | 2468 | ext4_msg(sb, KERN_ERR, "bad block size %d", |
ce40733c | 2469 | blocksize); |
ac27a0ec DK |
2470 | goto failed_mount; |
2471 | } | |
2472 | ||
2b2d6d01 | 2473 | brelse(bh); |
70bbb3e0 AM |
2474 | logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; |
2475 | offset = do_div(logical_sb_block, blocksize); | |
2476 | bh = sb_bread(sb, logical_sb_block); | |
ac27a0ec | 2477 | if (!bh) { |
b31e1552 ES |
2478 | ext4_msg(sb, KERN_ERR, |
2479 | "Can't read superblock on 2nd try"); | |
ac27a0ec DK |
2480 | goto failed_mount; |
2481 | } | |
617ba13b | 2482 | es = (struct ext4_super_block *)(((char *)bh->b_data) + offset); |
ac27a0ec | 2483 | sbi->s_es = es; |
617ba13b | 2484 | if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { |
b31e1552 ES |
2485 | ext4_msg(sb, KERN_ERR, |
2486 | "Magic mismatch, very weird!"); | |
ac27a0ec DK |
2487 | goto failed_mount; |
2488 | } | |
2489 | } | |
2490 | ||
a13fb1a4 ES |
2491 | has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb, |
2492 | EXT4_FEATURE_RO_COMPAT_HUGE_FILE); | |
f287a1a5 TT |
2493 | sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, |
2494 | has_huge_files); | |
2495 | sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); | |
ac27a0ec | 2496 | |
617ba13b MC |
2497 | if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { |
2498 | sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; | |
2499 | sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; | |
ac27a0ec DK |
2500 | } else { |
2501 | sbi->s_inode_size = le16_to_cpu(es->s_inode_size); | |
2502 | sbi->s_first_ino = le32_to_cpu(es->s_first_ino); | |
617ba13b | 2503 | if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || |
1330593e | 2504 | (!is_power_of_2(sbi->s_inode_size)) || |
ac27a0ec | 2505 | (sbi->s_inode_size > blocksize)) { |
b31e1552 ES |
2506 | ext4_msg(sb, KERN_ERR, |
2507 | "unsupported inode size: %d", | |
2b2d6d01 | 2508 | sbi->s_inode_size); |
ac27a0ec DK |
2509 | goto failed_mount; |
2510 | } | |
ef7f3835 KS |
2511 | if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) |
2512 | sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2); | |
ac27a0ec | 2513 | } |
0b8e58a1 | 2514 | |
0d1ee42f AR |
2515 | sbi->s_desc_size = le16_to_cpu(es->s_desc_size); |
2516 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) { | |
8fadc143 | 2517 | if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || |
0d1ee42f | 2518 | sbi->s_desc_size > EXT4_MAX_DESC_SIZE || |
d8ea6cf8 | 2519 | !is_power_of_2(sbi->s_desc_size)) { |
b31e1552 ES |
2520 | ext4_msg(sb, KERN_ERR, |
2521 | "unsupported descriptor size %lu", | |
0d1ee42f AR |
2522 | sbi->s_desc_size); |
2523 | goto failed_mount; | |
2524 | } | |
2525 | } else | |
2526 | sbi->s_desc_size = EXT4_MIN_DESC_SIZE; | |
0b8e58a1 | 2527 | |
ac27a0ec | 2528 | sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); |
ac27a0ec | 2529 | sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); |
b47b6f38 | 2530 | if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0) |
617ba13b | 2531 | goto cantfind_ext4; |
0b8e58a1 | 2532 | |
617ba13b | 2533 | sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); |
ac27a0ec | 2534 | if (sbi->s_inodes_per_block == 0) |
617ba13b | 2535 | goto cantfind_ext4; |
ac27a0ec DK |
2536 | sbi->s_itb_per_group = sbi->s_inodes_per_group / |
2537 | sbi->s_inodes_per_block; | |
0d1ee42f | 2538 | sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); |
ac27a0ec DK |
2539 | sbi->s_sbh = bh; |
2540 | sbi->s_mount_state = le16_to_cpu(es->s_state); | |
e57aa839 FW |
2541 | sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); |
2542 | sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); | |
0b8e58a1 | 2543 | |
2b2d6d01 | 2544 | for (i = 0; i < 4; i++) |
ac27a0ec DK |
2545 | sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); |
2546 | sbi->s_def_hash_version = es->s_def_hash_version; | |
f99b2589 TT |
2547 | i = le32_to_cpu(es->s_flags); |
2548 | if (i & EXT2_FLAGS_UNSIGNED_HASH) | |
2549 | sbi->s_hash_unsigned = 3; | |
2550 | else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { | |
2551 | #ifdef __CHAR_UNSIGNED__ | |
2552 | es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); | |
2553 | sbi->s_hash_unsigned = 3; | |
2554 | #else | |
2555 | es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); | |
2556 | #endif | |
2557 | sb->s_dirt = 1; | |
2558 | } | |
ac27a0ec DK |
2559 | |
2560 | if (sbi->s_blocks_per_group > blocksize * 8) { | |
b31e1552 ES |
2561 | ext4_msg(sb, KERN_ERR, |
2562 | "#blocks per group too big: %lu", | |
2b2d6d01 | 2563 | sbi->s_blocks_per_group); |
ac27a0ec DK |
2564 | goto failed_mount; |
2565 | } | |
ac27a0ec | 2566 | if (sbi->s_inodes_per_group > blocksize * 8) { |
b31e1552 ES |
2567 | ext4_msg(sb, KERN_ERR, |
2568 | "#inodes per group too big: %lu", | |
2b2d6d01 | 2569 | sbi->s_inodes_per_group); |
ac27a0ec DK |
2570 | goto failed_mount; |
2571 | } | |
2572 | ||
bf43d84b ES |
2573 | /* |
2574 | * Test whether we have more sectors than will fit in sector_t, | |
2575 | * and whether the max offset is addressable by the page cache. | |
2576 | */ | |
2577 | if ((ext4_blocks_count(es) > | |
2578 | (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) || | |
2579 | (ext4_blocks_count(es) > | |
2580 | (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) { | |
b31e1552 | 2581 | ext4_msg(sb, KERN_ERR, "filesystem" |
bf43d84b | 2582 | " too large to mount safely on this system"); |
ac27a0ec | 2583 | if (sizeof(sector_t) < 8) |
90c699a9 | 2584 | ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled"); |
bf43d84b | 2585 | ret = -EFBIG; |
ac27a0ec DK |
2586 | goto failed_mount; |
2587 | } | |
2588 | ||
617ba13b MC |
2589 | if (EXT4_BLOCKS_PER_GROUP(sb) == 0) |
2590 | goto cantfind_ext4; | |
e7c95593 | 2591 | |
0f2ddca6 FTN |
2592 | /* check blocks count against device size */ |
2593 | blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; | |
2594 | if (blocks_count && ext4_blocks_count(es) > blocks_count) { | |
b31e1552 ES |
2595 | ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " |
2596 | "exceeds size of device (%llu blocks)", | |
0f2ddca6 FTN |
2597 | ext4_blocks_count(es), blocks_count); |
2598 | goto failed_mount; | |
2599 | } | |
2600 | ||
0b8e58a1 AD |
2601 | /* |
2602 | * It makes no sense for the first data block to be beyond the end | |
2603 | * of the filesystem. | |
2604 | */ | |
2605 | if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { | |
b31e1552 ES |
2606 | ext4_msg(sb, KERN_WARNING, "bad geometry: first data" |
2607 | "block %u is beyond end of filesystem (%llu)", | |
2608 | le32_to_cpu(es->s_first_data_block), | |
2609 | ext4_blocks_count(es)); | |
e7c95593 ES |
2610 | goto failed_mount; |
2611 | } | |
bd81d8ee LV |
2612 | blocks_count = (ext4_blocks_count(es) - |
2613 | le32_to_cpu(es->s_first_data_block) + | |
2614 | EXT4_BLOCKS_PER_GROUP(sb) - 1); | |
2615 | do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); | |
4ec11028 | 2616 | if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { |
b31e1552 | 2617 | ext4_msg(sb, KERN_WARNING, "groups count too large: %u " |
4ec11028 | 2618 | "(block count %llu, first data block %u, " |
b31e1552 | 2619 | "blocks per group %lu)", sbi->s_groups_count, |
4ec11028 TT |
2620 | ext4_blocks_count(es), |
2621 | le32_to_cpu(es->s_first_data_block), | |
2622 | EXT4_BLOCKS_PER_GROUP(sb)); | |
2623 | goto failed_mount; | |
2624 | } | |
bd81d8ee | 2625 | sbi->s_groups_count = blocks_count; |
fb0a387d ES |
2626 | sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, |
2627 | (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); | |
617ba13b MC |
2628 | db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / |
2629 | EXT4_DESC_PER_BLOCK(sb); | |
2b2d6d01 | 2630 | sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *), |
ac27a0ec DK |
2631 | GFP_KERNEL); |
2632 | if (sbi->s_group_desc == NULL) { | |
b31e1552 | 2633 | ext4_msg(sb, KERN_ERR, "not enough memory"); |
ac27a0ec DK |
2634 | goto failed_mount; |
2635 | } | |
2636 | ||
3244fcb1 | 2637 | #ifdef CONFIG_PROC_FS |
9f6200bb TT |
2638 | if (ext4_proc_root) |
2639 | sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root); | |
3244fcb1 | 2640 | #endif |
240799cd | 2641 | |
705895b6 | 2642 | bgl_lock_init(sbi->s_blockgroup_lock); |
ac27a0ec DK |
2643 | |
2644 | for (i = 0; i < db_count; i++) { | |
70bbb3e0 | 2645 | block = descriptor_loc(sb, logical_sb_block, i); |
ac27a0ec DK |
2646 | sbi->s_group_desc[i] = sb_bread(sb, block); |
2647 | if (!sbi->s_group_desc[i]) { | |
b31e1552 ES |
2648 | ext4_msg(sb, KERN_ERR, |
2649 | "can't read group descriptor %d", i); | |
ac27a0ec DK |
2650 | db_count = i; |
2651 | goto failed_mount2; | |
2652 | } | |
2653 | } | |
2b2d6d01 | 2654 | if (!ext4_check_descriptors(sb)) { |
b31e1552 | 2655 | ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); |
ac27a0ec DK |
2656 | goto failed_mount2; |
2657 | } | |
772cb7c8 JS |
2658 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) |
2659 | if (!ext4_fill_flex_info(sb)) { | |
b31e1552 ES |
2660 | ext4_msg(sb, KERN_ERR, |
2661 | "unable to initialize " | |
2662 | "flex_bg meta info!"); | |
772cb7c8 JS |
2663 | goto failed_mount2; |
2664 | } | |
2665 | ||
ac27a0ec DK |
2666 | sbi->s_gdb_count = db_count; |
2667 | get_random_bytes(&sbi->s_next_generation, sizeof(u32)); | |
2668 | spin_lock_init(&sbi->s_next_gen_lock); | |
2669 | ||
833f4077 PZ |
2670 | err = percpu_counter_init(&sbi->s_freeblocks_counter, |
2671 | ext4_count_free_blocks(sb)); | |
2672 | if (!err) { | |
2673 | err = percpu_counter_init(&sbi->s_freeinodes_counter, | |
2674 | ext4_count_free_inodes(sb)); | |
2675 | } | |
2676 | if (!err) { | |
2677 | err = percpu_counter_init(&sbi->s_dirs_counter, | |
2678 | ext4_count_dirs(sb)); | |
2679 | } | |
6bc6e63f AK |
2680 | if (!err) { |
2681 | err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0); | |
2682 | } | |
833f4077 | 2683 | if (err) { |
b31e1552 | 2684 | ext4_msg(sb, KERN_ERR, "insufficient memory"); |
833f4077 PZ |
2685 | goto failed_mount3; |
2686 | } | |
ac27a0ec | 2687 | |
c9de560d | 2688 | sbi->s_stripe = ext4_get_stripe_size(sbi); |
55138e0b | 2689 | sbi->s_max_writeback_mb_bump = 128; |
c9de560d | 2690 | |
ac27a0ec DK |
2691 | /* |
2692 | * set up enough so that it can read an inode | |
2693 | */ | |
9ca92389 TT |
2694 | if (!test_opt(sb, NOLOAD) && |
2695 | EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) | |
2696 | sb->s_op = &ext4_sops; | |
2697 | else | |
2698 | sb->s_op = &ext4_nojournal_sops; | |
617ba13b MC |
2699 | sb->s_export_op = &ext4_export_ops; |
2700 | sb->s_xattr = ext4_xattr_handlers; | |
ac27a0ec | 2701 | #ifdef CONFIG_QUOTA |
617ba13b MC |
2702 | sb->s_qcop = &ext4_qctl_operations; |
2703 | sb->dq_op = &ext4_quota_operations; | |
ac27a0ec DK |
2704 | #endif |
2705 | INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ | |
3b9d4ed2 | 2706 | mutex_init(&sbi->s_orphan_lock); |
32ed5058 | 2707 | mutex_init(&sbi->s_resize_lock); |
ac27a0ec DK |
2708 | |
2709 | sb->s_root = NULL; | |
2710 | ||
2711 | needs_recovery = (es->s_last_orphan != 0 || | |
617ba13b MC |
2712 | EXT4_HAS_INCOMPAT_FEATURE(sb, |
2713 | EXT4_FEATURE_INCOMPAT_RECOVER)); | |
ac27a0ec DK |
2714 | |
2715 | /* | |
2716 | * The first inode we look at is the journal inode. Don't try | |
2717 | * root first: it may be modified in the journal! | |
2718 | */ | |
2719 | if (!test_opt(sb, NOLOAD) && | |
617ba13b MC |
2720 | EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { |
2721 | if (ext4_load_journal(sb, es, journal_devnum)) | |
ac27a0ec | 2722 | goto failed_mount3; |
624080ed TT |
2723 | if (!(sb->s_flags & MS_RDONLY) && |
2724 | EXT4_SB(sb)->s_journal->j_failed_commit) { | |
b31e1552 | 2725 | ext4_msg(sb, KERN_CRIT, "error: " |
624080ed | 2726 | "ext4_fill_super: Journal transaction " |
b31e1552 | 2727 | "%u is corrupt", |
624080ed | 2728 | EXT4_SB(sb)->s_journal->j_failed_commit); |
2b2d6d01 | 2729 | if (test_opt(sb, ERRORS_RO)) { |
b31e1552 ES |
2730 | ext4_msg(sb, KERN_CRIT, |
2731 | "Mounting filesystem read-only"); | |
624080ed TT |
2732 | sb->s_flags |= MS_RDONLY; |
2733 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; | |
2734 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
2735 | } | |
2736 | if (test_opt(sb, ERRORS_PANIC)) { | |
2737 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; | |
2738 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
e2d67052 | 2739 | ext4_commit_super(sb, 1); |
624080ed TT |
2740 | goto failed_mount4; |
2741 | } | |
2742 | } | |
0390131b FM |
2743 | } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) && |
2744 | EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { | |
b31e1552 ES |
2745 | ext4_msg(sb, KERN_ERR, "required journal recovery " |
2746 | "suppressed and not mounted read-only"); | |
0390131b | 2747 | goto failed_mount4; |
ac27a0ec | 2748 | } else { |
0390131b FM |
2749 | clear_opt(sbi->s_mount_opt, DATA_FLAGS); |
2750 | set_opt(sbi->s_mount_opt, WRITEBACK_DATA); | |
2751 | sbi->s_journal = NULL; | |
2752 | needs_recovery = 0; | |
2753 | goto no_journal; | |
ac27a0ec DK |
2754 | } |
2755 | ||
eb40a09c JS |
2756 | if (ext4_blocks_count(es) > 0xffffffffULL && |
2757 | !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, | |
2758 | JBD2_FEATURE_INCOMPAT_64BIT)) { | |
b31e1552 | 2759 | ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); |
eb40a09c JS |
2760 | goto failed_mount4; |
2761 | } | |
2762 | ||
d0646f7b TT |
2763 | jbd2_journal_set_features(sbi->s_journal, |
2764 | JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0); | |
2765 | if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) | |
2766 | jbd2_journal_set_features(sbi->s_journal, 0, 0, | |
818d276c | 2767 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); |
d0646f7b | 2768 | else |
818d276c GS |
2769 | jbd2_journal_clear_features(sbi->s_journal, 0, 0, |
2770 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); | |
818d276c | 2771 | |
ac27a0ec DK |
2772 | /* We have now updated the journal if required, so we can |
2773 | * validate the data journaling mode. */ | |
2774 | switch (test_opt(sb, DATA_FLAGS)) { | |
2775 | case 0: | |
2776 | /* No mode set, assume a default based on the journal | |
63f57933 AM |
2777 | * capabilities: ORDERED_DATA if the journal can |
2778 | * cope, else JOURNAL_DATA | |
2779 | */ | |
dab291af MC |
2780 | if (jbd2_journal_check_available_features |
2781 | (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) | |
ac27a0ec DK |
2782 | set_opt(sbi->s_mount_opt, ORDERED_DATA); |
2783 | else | |
2784 | set_opt(sbi->s_mount_opt, JOURNAL_DATA); | |
2785 | break; | |
2786 | ||
617ba13b MC |
2787 | case EXT4_MOUNT_ORDERED_DATA: |
2788 | case EXT4_MOUNT_WRITEBACK_DATA: | |
dab291af MC |
2789 | if (!jbd2_journal_check_available_features |
2790 | (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { | |
b31e1552 ES |
2791 | ext4_msg(sb, KERN_ERR, "Journal does not support " |
2792 | "requested data journaling mode"); | |
ac27a0ec DK |
2793 | goto failed_mount4; |
2794 | } | |
2795 | default: | |
2796 | break; | |
2797 | } | |
b3881f74 | 2798 | set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); |
ac27a0ec | 2799 | |
0390131b | 2800 | no_journal: |
ac27a0ec DK |
2801 | |
2802 | if (test_opt(sb, NOBH)) { | |
617ba13b | 2803 | if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) { |
b31e1552 ES |
2804 | ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - " |
2805 | "its supported only with writeback mode"); | |
ac27a0ec DK |
2806 | clear_opt(sbi->s_mount_opt, NOBH); |
2807 | } | |
2808 | } | |
4c0425ff MC |
2809 | EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten"); |
2810 | if (!EXT4_SB(sb)->dio_unwritten_wq) { | |
2811 | printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n"); | |
2812 | goto failed_mount_wq; | |
2813 | } | |
2814 | ||
ac27a0ec | 2815 | /* |
dab291af | 2816 | * The jbd2_journal_load will have done any necessary log recovery, |
ac27a0ec DK |
2817 | * so we can safely mount the rest of the filesystem now. |
2818 | */ | |
2819 | ||
1d1fe1ee DH |
2820 | root = ext4_iget(sb, EXT4_ROOT_INO); |
2821 | if (IS_ERR(root)) { | |
b31e1552 | 2822 | ext4_msg(sb, KERN_ERR, "get root inode failed"); |
1d1fe1ee | 2823 | ret = PTR_ERR(root); |
ac27a0ec DK |
2824 | goto failed_mount4; |
2825 | } | |
2826 | if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { | |
1d1fe1ee | 2827 | iput(root); |
b31e1552 | 2828 | ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); |
ac27a0ec DK |
2829 | goto failed_mount4; |
2830 | } | |
1d1fe1ee DH |
2831 | sb->s_root = d_alloc_root(root); |
2832 | if (!sb->s_root) { | |
b31e1552 | 2833 | ext4_msg(sb, KERN_ERR, "get root dentry failed"); |
1d1fe1ee DH |
2834 | iput(root); |
2835 | ret = -ENOMEM; | |
2836 | goto failed_mount4; | |
2837 | } | |
ac27a0ec | 2838 | |
2b2d6d01 | 2839 | ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY); |
ef7f3835 KS |
2840 | |
2841 | /* determine the minimum size of new large inodes, if present */ | |
2842 | if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { | |
2843 | sbi->s_want_extra_isize = sizeof(struct ext4_inode) - | |
2844 | EXT4_GOOD_OLD_INODE_SIZE; | |
2845 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, | |
2846 | EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) { | |
2847 | if (sbi->s_want_extra_isize < | |
2848 | le16_to_cpu(es->s_want_extra_isize)) | |
2849 | sbi->s_want_extra_isize = | |
2850 | le16_to_cpu(es->s_want_extra_isize); | |
2851 | if (sbi->s_want_extra_isize < | |
2852 | le16_to_cpu(es->s_min_extra_isize)) | |
2853 | sbi->s_want_extra_isize = | |
2854 | le16_to_cpu(es->s_min_extra_isize); | |
2855 | } | |
2856 | } | |
2857 | /* Check if enough inode space is available */ | |
2858 | if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize > | |
2859 | sbi->s_inode_size) { | |
2860 | sbi->s_want_extra_isize = sizeof(struct ext4_inode) - | |
2861 | EXT4_GOOD_OLD_INODE_SIZE; | |
b31e1552 ES |
2862 | ext4_msg(sb, KERN_INFO, "required extra inode space not" |
2863 | "available"); | |
ef7f3835 KS |
2864 | } |
2865 | ||
c2774d84 | 2866 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { |
b31e1552 ES |
2867 | ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - " |
2868 | "requested data journaling mode"); | |
c2774d84 AK |
2869 | clear_opt(sbi->s_mount_opt, DELALLOC); |
2870 | } else if (test_opt(sb, DELALLOC)) | |
b31e1552 | 2871 | ext4_msg(sb, KERN_INFO, "delayed allocation enabled"); |
c2774d84 | 2872 | |
6fd058f7 TT |
2873 | err = ext4_setup_system_zone(sb); |
2874 | if (err) { | |
b31e1552 ES |
2875 | ext4_msg(sb, KERN_ERR, "failed to initialize system " |
2876 | "zone (%d)\n", err); | |
6fd058f7 TT |
2877 | goto failed_mount4; |
2878 | } | |
2879 | ||
c2774d84 AK |
2880 | ext4_ext_init(sb); |
2881 | err = ext4_mb_init(sb, needs_recovery); | |
2882 | if (err) { | |
b31e1552 ES |
2883 | ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)", |
2884 | err); | |
c2774d84 AK |
2885 | goto failed_mount4; |
2886 | } | |
2887 | ||
3197ebdb TT |
2888 | sbi->s_kobj.kset = ext4_kset; |
2889 | init_completion(&sbi->s_kobj_unregister); | |
2890 | err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL, | |
2891 | "%s", sb->s_id); | |
2892 | if (err) { | |
2893 | ext4_mb_release(sb); | |
2894 | ext4_ext_release(sb); | |
2895 | goto failed_mount4; | |
2896 | }; | |
2897 | ||
617ba13b MC |
2898 | EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; |
2899 | ext4_orphan_cleanup(sb, es); | |
2900 | EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; | |
0390131b | 2901 | if (needs_recovery) { |
b31e1552 | 2902 | ext4_msg(sb, KERN_INFO, "recovery complete"); |
0390131b FM |
2903 | ext4_mark_recovery_complete(sb, es); |
2904 | } | |
2905 | if (EXT4_SB(sb)->s_journal) { | |
2906 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) | |
2907 | descr = " journalled data mode"; | |
2908 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) | |
2909 | descr = " ordered data mode"; | |
2910 | else | |
2911 | descr = " writeback data mode"; | |
2912 | } else | |
2913 | descr = "out journal"; | |
2914 | ||
b31e1552 | 2915 | ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr); |
ac27a0ec DK |
2916 | |
2917 | lock_kernel(); | |
2918 | return 0; | |
2919 | ||
617ba13b | 2920 | cantfind_ext4: |
ac27a0ec | 2921 | if (!silent) |
b31e1552 | 2922 | ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); |
ac27a0ec DK |
2923 | goto failed_mount; |
2924 | ||
2925 | failed_mount4: | |
b31e1552 | 2926 | ext4_msg(sb, KERN_ERR, "mount failed"); |
4c0425ff MC |
2927 | destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq); |
2928 | failed_mount_wq: | |
6fd058f7 | 2929 | ext4_release_system_zone(sb); |
0390131b FM |
2930 | if (sbi->s_journal) { |
2931 | jbd2_journal_destroy(sbi->s_journal); | |
2932 | sbi->s_journal = NULL; | |
2933 | } | |
ac27a0ec | 2934 | failed_mount3: |
c5ca7c76 TT |
2935 | if (sbi->s_flex_groups) { |
2936 | if (is_vmalloc_addr(sbi->s_flex_groups)) | |
2937 | vfree(sbi->s_flex_groups); | |
2938 | else | |
2939 | kfree(sbi->s_flex_groups); | |
2940 | } | |
ac27a0ec DK |
2941 | percpu_counter_destroy(&sbi->s_freeblocks_counter); |
2942 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
2943 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
6bc6e63f | 2944 | percpu_counter_destroy(&sbi->s_dirtyblocks_counter); |
ac27a0ec DK |
2945 | failed_mount2: |
2946 | for (i = 0; i < db_count; i++) | |
2947 | brelse(sbi->s_group_desc[i]); | |
2948 | kfree(sbi->s_group_desc); | |
2949 | failed_mount: | |
240799cd | 2950 | if (sbi->s_proc) { |
9f6200bb | 2951 | remove_proc_entry(sb->s_id, ext4_proc_root); |
240799cd | 2952 | } |
ac27a0ec DK |
2953 | #ifdef CONFIG_QUOTA |
2954 | for (i = 0; i < MAXQUOTAS; i++) | |
2955 | kfree(sbi->s_qf_names[i]); | |
2956 | #endif | |
617ba13b | 2957 | ext4_blkdev_remove(sbi); |
ac27a0ec DK |
2958 | brelse(bh); |
2959 | out_fail: | |
2960 | sb->s_fs_info = NULL; | |
f6830165 | 2961 | kfree(sbi->s_blockgroup_lock); |
ac27a0ec DK |
2962 | kfree(sbi); |
2963 | lock_kernel(); | |
1d1fe1ee | 2964 | return ret; |
ac27a0ec DK |
2965 | } |
2966 | ||
2967 | /* | |
2968 | * Setup any per-fs journal parameters now. We'll do this both on | |
2969 | * initial mount, once the journal has been initialised but before we've | |
2970 | * done any recovery; and again on any subsequent remount. | |
2971 | */ | |
617ba13b | 2972 | static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) |
ac27a0ec | 2973 | { |
617ba13b | 2974 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 2975 | |
30773840 TT |
2976 | journal->j_commit_interval = sbi->s_commit_interval; |
2977 | journal->j_min_batch_time = sbi->s_min_batch_time; | |
2978 | journal->j_max_batch_time = sbi->s_max_batch_time; | |
ac27a0ec DK |
2979 | |
2980 | spin_lock(&journal->j_state_lock); | |
2981 | if (test_opt(sb, BARRIER)) | |
dab291af | 2982 | journal->j_flags |= JBD2_BARRIER; |
ac27a0ec | 2983 | else |
dab291af | 2984 | journal->j_flags &= ~JBD2_BARRIER; |
5bf5683a HK |
2985 | if (test_opt(sb, DATA_ERR_ABORT)) |
2986 | journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; | |
2987 | else | |
2988 | journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; | |
ac27a0ec DK |
2989 | spin_unlock(&journal->j_state_lock); |
2990 | } | |
2991 | ||
617ba13b | 2992 | static journal_t *ext4_get_journal(struct super_block *sb, |
ac27a0ec DK |
2993 | unsigned int journal_inum) |
2994 | { | |
2995 | struct inode *journal_inode; | |
2996 | journal_t *journal; | |
2997 | ||
0390131b FM |
2998 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
2999 | ||
ac27a0ec DK |
3000 | /* First, test for the existence of a valid inode on disk. Bad |
3001 | * things happen if we iget() an unused inode, as the subsequent | |
3002 | * iput() will try to delete it. */ | |
3003 | ||
1d1fe1ee DH |
3004 | journal_inode = ext4_iget(sb, journal_inum); |
3005 | if (IS_ERR(journal_inode)) { | |
b31e1552 | 3006 | ext4_msg(sb, KERN_ERR, "no journal found"); |
ac27a0ec DK |
3007 | return NULL; |
3008 | } | |
3009 | if (!journal_inode->i_nlink) { | |
3010 | make_bad_inode(journal_inode); | |
3011 | iput(journal_inode); | |
b31e1552 | 3012 | ext4_msg(sb, KERN_ERR, "journal inode is deleted"); |
ac27a0ec DK |
3013 | return NULL; |
3014 | } | |
3015 | ||
e5f8eab8 | 3016 | jbd_debug(2, "Journal inode found at %p: %lld bytes\n", |
ac27a0ec | 3017 | journal_inode, journal_inode->i_size); |
1d1fe1ee | 3018 | if (!S_ISREG(journal_inode->i_mode)) { |
b31e1552 | 3019 | ext4_msg(sb, KERN_ERR, "invalid journal inode"); |
ac27a0ec DK |
3020 | iput(journal_inode); |
3021 | return NULL; | |
3022 | } | |
3023 | ||
dab291af | 3024 | journal = jbd2_journal_init_inode(journal_inode); |
ac27a0ec | 3025 | if (!journal) { |
b31e1552 | 3026 | ext4_msg(sb, KERN_ERR, "Could not load journal inode"); |
ac27a0ec DK |
3027 | iput(journal_inode); |
3028 | return NULL; | |
3029 | } | |
3030 | journal->j_private = sb; | |
617ba13b | 3031 | ext4_init_journal_params(sb, journal); |
ac27a0ec DK |
3032 | return journal; |
3033 | } | |
3034 | ||
617ba13b | 3035 | static journal_t *ext4_get_dev_journal(struct super_block *sb, |
ac27a0ec DK |
3036 | dev_t j_dev) |
3037 | { | |
2b2d6d01 | 3038 | struct buffer_head *bh; |
ac27a0ec | 3039 | journal_t *journal; |
617ba13b MC |
3040 | ext4_fsblk_t start; |
3041 | ext4_fsblk_t len; | |
ac27a0ec | 3042 | int hblock, blocksize; |
617ba13b | 3043 | ext4_fsblk_t sb_block; |
ac27a0ec | 3044 | unsigned long offset; |
2b2d6d01 | 3045 | struct ext4_super_block *es; |
ac27a0ec DK |
3046 | struct block_device *bdev; |
3047 | ||
0390131b FM |
3048 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
3049 | ||
b31e1552 | 3050 | bdev = ext4_blkdev_get(j_dev, sb); |
ac27a0ec DK |
3051 | if (bdev == NULL) |
3052 | return NULL; | |
3053 | ||
3054 | if (bd_claim(bdev, sb)) { | |
b31e1552 ES |
3055 | ext4_msg(sb, KERN_ERR, |
3056 | "failed to claim external journal device"); | |
9a1c3542 | 3057 | blkdev_put(bdev, FMODE_READ|FMODE_WRITE); |
ac27a0ec DK |
3058 | return NULL; |
3059 | } | |
3060 | ||
3061 | blocksize = sb->s_blocksize; | |
e1defc4f | 3062 | hblock = bdev_logical_block_size(bdev); |
ac27a0ec | 3063 | if (blocksize < hblock) { |
b31e1552 ES |
3064 | ext4_msg(sb, KERN_ERR, |
3065 | "blocksize too small for journal device"); | |
ac27a0ec DK |
3066 | goto out_bdev; |
3067 | } | |
3068 | ||
617ba13b MC |
3069 | sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; |
3070 | offset = EXT4_MIN_BLOCK_SIZE % blocksize; | |
ac27a0ec DK |
3071 | set_blocksize(bdev, blocksize); |
3072 | if (!(bh = __bread(bdev, sb_block, blocksize))) { | |
b31e1552 ES |
3073 | ext4_msg(sb, KERN_ERR, "couldn't read superblock of " |
3074 | "external journal"); | |
ac27a0ec DK |
3075 | goto out_bdev; |
3076 | } | |
3077 | ||
617ba13b MC |
3078 | es = (struct ext4_super_block *) (((char *)bh->b_data) + offset); |
3079 | if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || | |
ac27a0ec | 3080 | !(le32_to_cpu(es->s_feature_incompat) & |
617ba13b | 3081 | EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { |
b31e1552 ES |
3082 | ext4_msg(sb, KERN_ERR, "external journal has " |
3083 | "bad superblock"); | |
ac27a0ec DK |
3084 | brelse(bh); |
3085 | goto out_bdev; | |
3086 | } | |
3087 | ||
617ba13b | 3088 | if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { |
b31e1552 | 3089 | ext4_msg(sb, KERN_ERR, "journal UUID does not match"); |
ac27a0ec DK |
3090 | brelse(bh); |
3091 | goto out_bdev; | |
3092 | } | |
3093 | ||
bd81d8ee | 3094 | len = ext4_blocks_count(es); |
ac27a0ec DK |
3095 | start = sb_block + 1; |
3096 | brelse(bh); /* we're done with the superblock */ | |
3097 | ||
dab291af | 3098 | journal = jbd2_journal_init_dev(bdev, sb->s_bdev, |
ac27a0ec DK |
3099 | start, len, blocksize); |
3100 | if (!journal) { | |
b31e1552 | 3101 | ext4_msg(sb, KERN_ERR, "failed to create device journal"); |
ac27a0ec DK |
3102 | goto out_bdev; |
3103 | } | |
3104 | journal->j_private = sb; | |
3105 | ll_rw_block(READ, 1, &journal->j_sb_buffer); | |
3106 | wait_on_buffer(journal->j_sb_buffer); | |
3107 | if (!buffer_uptodate(journal->j_sb_buffer)) { | |
b31e1552 | 3108 | ext4_msg(sb, KERN_ERR, "I/O error on journal device"); |
ac27a0ec DK |
3109 | goto out_journal; |
3110 | } | |
3111 | if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { | |
b31e1552 ES |
3112 | ext4_msg(sb, KERN_ERR, "External journal has more than one " |
3113 | "user (unsupported) - %d", | |
ac27a0ec DK |
3114 | be32_to_cpu(journal->j_superblock->s_nr_users)); |
3115 | goto out_journal; | |
3116 | } | |
617ba13b MC |
3117 | EXT4_SB(sb)->journal_bdev = bdev; |
3118 | ext4_init_journal_params(sb, journal); | |
ac27a0ec | 3119 | return journal; |
0b8e58a1 | 3120 | |
ac27a0ec | 3121 | out_journal: |
dab291af | 3122 | jbd2_journal_destroy(journal); |
ac27a0ec | 3123 | out_bdev: |
617ba13b | 3124 | ext4_blkdev_put(bdev); |
ac27a0ec DK |
3125 | return NULL; |
3126 | } | |
3127 | ||
617ba13b MC |
3128 | static int ext4_load_journal(struct super_block *sb, |
3129 | struct ext4_super_block *es, | |
ac27a0ec DK |
3130 | unsigned long journal_devnum) |
3131 | { | |
3132 | journal_t *journal; | |
3133 | unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); | |
3134 | dev_t journal_dev; | |
3135 | int err = 0; | |
3136 | int really_read_only; | |
3137 | ||
0390131b FM |
3138 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
3139 | ||
ac27a0ec DK |
3140 | if (journal_devnum && |
3141 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { | |
b31e1552 ES |
3142 | ext4_msg(sb, KERN_INFO, "external journal device major/minor " |
3143 | "numbers have changed"); | |
ac27a0ec DK |
3144 | journal_dev = new_decode_dev(journal_devnum); |
3145 | } else | |
3146 | journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); | |
3147 | ||
3148 | really_read_only = bdev_read_only(sb->s_bdev); | |
3149 | ||
3150 | /* | |
3151 | * Are we loading a blank journal or performing recovery after a | |
3152 | * crash? For recovery, we need to check in advance whether we | |
3153 | * can get read-write access to the device. | |
3154 | */ | |
617ba13b | 3155 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { |
ac27a0ec | 3156 | if (sb->s_flags & MS_RDONLY) { |
b31e1552 ES |
3157 | ext4_msg(sb, KERN_INFO, "INFO: recovery " |
3158 | "required on readonly filesystem"); | |
ac27a0ec | 3159 | if (really_read_only) { |
b31e1552 ES |
3160 | ext4_msg(sb, KERN_ERR, "write access " |
3161 | "unavailable, cannot proceed"); | |
ac27a0ec DK |
3162 | return -EROFS; |
3163 | } | |
b31e1552 ES |
3164 | ext4_msg(sb, KERN_INFO, "write access will " |
3165 | "be enabled during recovery"); | |
ac27a0ec DK |
3166 | } |
3167 | } | |
3168 | ||
3169 | if (journal_inum && journal_dev) { | |
b31e1552 ES |
3170 | ext4_msg(sb, KERN_ERR, "filesystem has both journal " |
3171 | "and inode journals!"); | |
ac27a0ec DK |
3172 | return -EINVAL; |
3173 | } | |
3174 | ||
3175 | if (journal_inum) { | |
617ba13b | 3176 | if (!(journal = ext4_get_journal(sb, journal_inum))) |
ac27a0ec DK |
3177 | return -EINVAL; |
3178 | } else { | |
617ba13b | 3179 | if (!(journal = ext4_get_dev_journal(sb, journal_dev))) |
ac27a0ec DK |
3180 | return -EINVAL; |
3181 | } | |
3182 | ||
4776004f | 3183 | if (journal->j_flags & JBD2_BARRIER) |
b31e1552 | 3184 | ext4_msg(sb, KERN_INFO, "barriers enabled"); |
4776004f | 3185 | else |
b31e1552 | 3186 | ext4_msg(sb, KERN_INFO, "barriers disabled"); |
4776004f | 3187 | |
ac27a0ec | 3188 | if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) { |
dab291af | 3189 | err = jbd2_journal_update_format(journal); |
ac27a0ec | 3190 | if (err) { |
b31e1552 | 3191 | ext4_msg(sb, KERN_ERR, "error updating journal"); |
dab291af | 3192 | jbd2_journal_destroy(journal); |
ac27a0ec DK |
3193 | return err; |
3194 | } | |
3195 | } | |
3196 | ||
617ba13b | 3197 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) |
dab291af | 3198 | err = jbd2_journal_wipe(journal, !really_read_only); |
ac27a0ec | 3199 | if (!err) |
dab291af | 3200 | err = jbd2_journal_load(journal); |
ac27a0ec DK |
3201 | |
3202 | if (err) { | |
b31e1552 | 3203 | ext4_msg(sb, KERN_ERR, "error loading journal"); |
dab291af | 3204 | jbd2_journal_destroy(journal); |
ac27a0ec DK |
3205 | return err; |
3206 | } | |
3207 | ||
617ba13b MC |
3208 | EXT4_SB(sb)->s_journal = journal; |
3209 | ext4_clear_journal_err(sb, es); | |
ac27a0ec DK |
3210 | |
3211 | if (journal_devnum && | |
3212 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { | |
3213 | es->s_journal_dev = cpu_to_le32(journal_devnum); | |
ac27a0ec DK |
3214 | |
3215 | /* Make sure we flush the recovery flag to disk. */ | |
e2d67052 | 3216 | ext4_commit_super(sb, 1); |
ac27a0ec DK |
3217 | } |
3218 | ||
3219 | return 0; | |
3220 | } | |
3221 | ||
e2d67052 | 3222 | static int ext4_commit_super(struct super_block *sb, int sync) |
ac27a0ec | 3223 | { |
e2d67052 | 3224 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
617ba13b | 3225 | struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; |
c4be0c1d | 3226 | int error = 0; |
ac27a0ec DK |
3227 | |
3228 | if (!sbh) | |
c4be0c1d | 3229 | return error; |
914258bf TT |
3230 | if (buffer_write_io_error(sbh)) { |
3231 | /* | |
3232 | * Oh, dear. A previous attempt to write the | |
3233 | * superblock failed. This could happen because the | |
3234 | * USB device was yanked out. Or it could happen to | |
3235 | * be a transient write error and maybe the block will | |
3236 | * be remapped. Nothing we can do but to retry the | |
3237 | * write and hope for the best. | |
3238 | */ | |
b31e1552 ES |
3239 | ext4_msg(sb, KERN_ERR, "previous I/O error to " |
3240 | "superblock detected"); | |
914258bf TT |
3241 | clear_buffer_write_io_error(sbh); |
3242 | set_buffer_uptodate(sbh); | |
3243 | } | |
71290b36 TT |
3244 | /* |
3245 | * If the file system is mounted read-only, don't update the | |
3246 | * superblock write time. This avoids updating the superblock | |
3247 | * write time when we are mounting the root file system | |
3248 | * read/only but we need to replay the journal; at that point, | |
3249 | * for people who are east of GMT and who make their clock | |
3250 | * tick in localtime for Windows bug-for-bug compatibility, | |
3251 | * the clock is set in the future, and this will cause e2fsck | |
3252 | * to complain and force a full file system check. | |
3253 | */ | |
3254 | if (!(sb->s_flags & MS_RDONLY)) | |
3255 | es->s_wtime = cpu_to_le32(get_seconds()); | |
afc32f7e TT |
3256 | es->s_kbytes_written = |
3257 | cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + | |
3258 | ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - | |
3259 | EXT4_SB(sb)->s_sectors_written_start) >> 1)); | |
5d1b1b3f AK |
3260 | ext4_free_blocks_count_set(es, percpu_counter_sum_positive( |
3261 | &EXT4_SB(sb)->s_freeblocks_counter)); | |
3262 | es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive( | |
3263 | &EXT4_SB(sb)->s_freeinodes_counter)); | |
7234ab2a | 3264 | sb->s_dirt = 0; |
ac27a0ec DK |
3265 | BUFFER_TRACE(sbh, "marking dirty"); |
3266 | mark_buffer_dirty(sbh); | |
914258bf | 3267 | if (sync) { |
c4be0c1d TS |
3268 | error = sync_dirty_buffer(sbh); |
3269 | if (error) | |
3270 | return error; | |
3271 | ||
3272 | error = buffer_write_io_error(sbh); | |
3273 | if (error) { | |
b31e1552 ES |
3274 | ext4_msg(sb, KERN_ERR, "I/O error while writing " |
3275 | "superblock"); | |
914258bf TT |
3276 | clear_buffer_write_io_error(sbh); |
3277 | set_buffer_uptodate(sbh); | |
3278 | } | |
3279 | } | |
c4be0c1d | 3280 | return error; |
ac27a0ec DK |
3281 | } |
3282 | ||
ac27a0ec DK |
3283 | /* |
3284 | * Have we just finished recovery? If so, and if we are mounting (or | |
3285 | * remounting) the filesystem readonly, then we will end up with a | |
3286 | * consistent fs on disk. Record that fact. | |
3287 | */ | |
2b2d6d01 TT |
3288 | static void ext4_mark_recovery_complete(struct super_block *sb, |
3289 | struct ext4_super_block *es) | |
ac27a0ec | 3290 | { |
617ba13b | 3291 | journal_t *journal = EXT4_SB(sb)->s_journal; |
ac27a0ec | 3292 | |
0390131b FM |
3293 | if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { |
3294 | BUG_ON(journal != NULL); | |
3295 | return; | |
3296 | } | |
dab291af | 3297 | jbd2_journal_lock_updates(journal); |
7ffe1ea8 HK |
3298 | if (jbd2_journal_flush(journal) < 0) |
3299 | goto out; | |
3300 | ||
617ba13b | 3301 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) && |
ac27a0ec | 3302 | sb->s_flags & MS_RDONLY) { |
617ba13b | 3303 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); |
e2d67052 | 3304 | ext4_commit_super(sb, 1); |
ac27a0ec | 3305 | } |
7ffe1ea8 HK |
3306 | |
3307 | out: | |
dab291af | 3308 | jbd2_journal_unlock_updates(journal); |
ac27a0ec DK |
3309 | } |
3310 | ||
3311 | /* | |
3312 | * If we are mounting (or read-write remounting) a filesystem whose journal | |
3313 | * has recorded an error from a previous lifetime, move that error to the | |
3314 | * main filesystem now. | |
3315 | */ | |
2b2d6d01 TT |
3316 | static void ext4_clear_journal_err(struct super_block *sb, |
3317 | struct ext4_super_block *es) | |
ac27a0ec DK |
3318 | { |
3319 | journal_t *journal; | |
3320 | int j_errno; | |
3321 | const char *errstr; | |
3322 | ||
0390131b FM |
3323 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
3324 | ||
617ba13b | 3325 | journal = EXT4_SB(sb)->s_journal; |
ac27a0ec DK |
3326 | |
3327 | /* | |
3328 | * Now check for any error status which may have been recorded in the | |
617ba13b | 3329 | * journal by a prior ext4_error() or ext4_abort() |
ac27a0ec DK |
3330 | */ |
3331 | ||
dab291af | 3332 | j_errno = jbd2_journal_errno(journal); |
ac27a0ec DK |
3333 | if (j_errno) { |
3334 | char nbuf[16]; | |
3335 | ||
617ba13b | 3336 | errstr = ext4_decode_error(sb, j_errno, nbuf); |
46e665e9 | 3337 | ext4_warning(sb, __func__, "Filesystem error recorded " |
ac27a0ec | 3338 | "from previous mount: %s", errstr); |
46e665e9 | 3339 | ext4_warning(sb, __func__, "Marking fs in need of " |
ac27a0ec DK |
3340 | "filesystem check."); |
3341 | ||
617ba13b MC |
3342 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; |
3343 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
e2d67052 | 3344 | ext4_commit_super(sb, 1); |
ac27a0ec | 3345 | |
dab291af | 3346 | jbd2_journal_clear_err(journal); |
ac27a0ec DK |
3347 | } |
3348 | } | |
3349 | ||
3350 | /* | |
3351 | * Force the running and committing transactions to commit, | |
3352 | * and wait on the commit. | |
3353 | */ | |
617ba13b | 3354 | int ext4_force_commit(struct super_block *sb) |
ac27a0ec DK |
3355 | { |
3356 | journal_t *journal; | |
0390131b | 3357 | int ret = 0; |
ac27a0ec DK |
3358 | |
3359 | if (sb->s_flags & MS_RDONLY) | |
3360 | return 0; | |
3361 | ||
617ba13b | 3362 | journal = EXT4_SB(sb)->s_journal; |
7234ab2a | 3363 | if (journal) |
0390131b | 3364 | ret = ext4_journal_force_commit(journal); |
0390131b | 3365 | |
ac27a0ec DK |
3366 | return ret; |
3367 | } | |
3368 | ||
2b2d6d01 | 3369 | static void ext4_write_super(struct super_block *sb) |
ac27a0ec | 3370 | { |
ebc1ac16 | 3371 | lock_super(sb); |
9ca92389 | 3372 | ext4_commit_super(sb, 1); |
ebc1ac16 | 3373 | unlock_super(sb); |
ac27a0ec DK |
3374 | } |
3375 | ||
617ba13b | 3376 | static int ext4_sync_fs(struct super_block *sb, int wait) |
ac27a0ec | 3377 | { |
14ce0cb4 | 3378 | int ret = 0; |
9eddacf9 | 3379 | tid_t target; |
8d5d02e6 | 3380 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 3381 | |
9bffad1e | 3382 | trace_ext4_sync_fs(sb, wait); |
8d5d02e6 MC |
3383 | flush_workqueue(sbi->dio_unwritten_wq); |
3384 | if (jbd2_journal_start_commit(sbi->s_journal, &target)) { | |
9ca92389 | 3385 | if (wait) |
8d5d02e6 | 3386 | jbd2_log_wait_commit(sbi->s_journal, target); |
0390131b | 3387 | } |
14ce0cb4 | 3388 | return ret; |
ac27a0ec DK |
3389 | } |
3390 | ||
3391 | /* | |
3392 | * LVM calls this function before a (read-only) snapshot is created. This | |
3393 | * gives us a chance to flush the journal completely and mark the fs clean. | |
3394 | */ | |
c4be0c1d | 3395 | static int ext4_freeze(struct super_block *sb) |
ac27a0ec | 3396 | { |
c4be0c1d TS |
3397 | int error = 0; |
3398 | journal_t *journal; | |
ac27a0ec | 3399 | |
9ca92389 TT |
3400 | if (sb->s_flags & MS_RDONLY) |
3401 | return 0; | |
ac27a0ec | 3402 | |
9ca92389 | 3403 | journal = EXT4_SB(sb)->s_journal; |
7ffe1ea8 | 3404 | |
9ca92389 TT |
3405 | /* Now we set up the journal barrier. */ |
3406 | jbd2_journal_lock_updates(journal); | |
ac27a0ec | 3407 | |
9ca92389 TT |
3408 | /* |
3409 | * Don't clear the needs_recovery flag if we failed to flush | |
3410 | * the journal. | |
3411 | */ | |
3412 | error = jbd2_journal_flush(journal); | |
3413 | if (error < 0) { | |
3414 | out: | |
3415 | jbd2_journal_unlock_updates(journal); | |
3416 | return error; | |
ac27a0ec | 3417 | } |
9ca92389 TT |
3418 | |
3419 | /* Journal blocked and flushed, clear needs_recovery flag. */ | |
3420 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
3421 | error = ext4_commit_super(sb, 1); | |
3422 | if (error) | |
3423 | goto out; | |
c4be0c1d | 3424 | return 0; |
ac27a0ec DK |
3425 | } |
3426 | ||
3427 | /* | |
3428 | * Called by LVM after the snapshot is done. We need to reset the RECOVER | |
3429 | * flag here, even though the filesystem is not technically dirty yet. | |
3430 | */ | |
c4be0c1d | 3431 | static int ext4_unfreeze(struct super_block *sb) |
ac27a0ec | 3432 | { |
9ca92389 TT |
3433 | if (sb->s_flags & MS_RDONLY) |
3434 | return 0; | |
3435 | ||
3436 | lock_super(sb); | |
3437 | /* Reset the needs_recovery flag before the fs is unlocked. */ | |
3438 | EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
3439 | ext4_commit_super(sb, 1); | |
3440 | unlock_super(sb); | |
3441 | jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); | |
c4be0c1d | 3442 | return 0; |
ac27a0ec DK |
3443 | } |
3444 | ||
2b2d6d01 | 3445 | static int ext4_remount(struct super_block *sb, int *flags, char *data) |
ac27a0ec | 3446 | { |
2b2d6d01 | 3447 | struct ext4_super_block *es; |
617ba13b MC |
3448 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
3449 | ext4_fsblk_t n_blocks_count = 0; | |
ac27a0ec | 3450 | unsigned long old_sb_flags; |
617ba13b | 3451 | struct ext4_mount_options old_opts; |
8a266467 | 3452 | ext4_group_t g; |
b3881f74 | 3453 | unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; |
ac27a0ec DK |
3454 | int err; |
3455 | #ifdef CONFIG_QUOTA | |
3456 | int i; | |
3457 | #endif | |
3458 | ||
337eb00a AIB |
3459 | lock_kernel(); |
3460 | ||
ac27a0ec | 3461 | /* Store the original options */ |
bbd6851a | 3462 | lock_super(sb); |
ac27a0ec DK |
3463 | old_sb_flags = sb->s_flags; |
3464 | old_opts.s_mount_opt = sbi->s_mount_opt; | |
3465 | old_opts.s_resuid = sbi->s_resuid; | |
3466 | old_opts.s_resgid = sbi->s_resgid; | |
3467 | old_opts.s_commit_interval = sbi->s_commit_interval; | |
30773840 TT |
3468 | old_opts.s_min_batch_time = sbi->s_min_batch_time; |
3469 | old_opts.s_max_batch_time = sbi->s_max_batch_time; | |
ac27a0ec DK |
3470 | #ifdef CONFIG_QUOTA |
3471 | old_opts.s_jquota_fmt = sbi->s_jquota_fmt; | |
3472 | for (i = 0; i < MAXQUOTAS; i++) | |
3473 | old_opts.s_qf_names[i] = sbi->s_qf_names[i]; | |
3474 | #endif | |
b3881f74 TT |
3475 | if (sbi->s_journal && sbi->s_journal->j_task->io_context) |
3476 | journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; | |
ac27a0ec DK |
3477 | |
3478 | /* | |
3479 | * Allow the "check" option to be passed as a remount option. | |
3480 | */ | |
b3881f74 TT |
3481 | if (!parse_options(data, sb, NULL, &journal_ioprio, |
3482 | &n_blocks_count, 1)) { | |
ac27a0ec DK |
3483 | err = -EINVAL; |
3484 | goto restore_opts; | |
3485 | } | |
3486 | ||
4ab2f15b | 3487 | if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) |
46e665e9 | 3488 | ext4_abort(sb, __func__, "Abort forced by user"); |
ac27a0ec DK |
3489 | |
3490 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
617ba13b | 3491 | ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0); |
ac27a0ec DK |
3492 | |
3493 | es = sbi->s_es; | |
3494 | ||
b3881f74 | 3495 | if (sbi->s_journal) { |
0390131b | 3496 | ext4_init_journal_params(sb, sbi->s_journal); |
b3881f74 TT |
3497 | set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); |
3498 | } | |
ac27a0ec DK |
3499 | |
3500 | if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) || | |
bd81d8ee | 3501 | n_blocks_count > ext4_blocks_count(es)) { |
4ab2f15b | 3502 | if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { |
ac27a0ec DK |
3503 | err = -EROFS; |
3504 | goto restore_opts; | |
3505 | } | |
3506 | ||
3507 | if (*flags & MS_RDONLY) { | |
3508 | /* | |
3509 | * First of all, the unconditional stuff we have to do | |
3510 | * to disable replay of the journal when we next remount | |
3511 | */ | |
3512 | sb->s_flags |= MS_RDONLY; | |
3513 | ||
3514 | /* | |
3515 | * OK, test if we are remounting a valid rw partition | |
3516 | * readonly, and if so set the rdonly flag and then | |
3517 | * mark the partition as valid again. | |
3518 | */ | |
617ba13b MC |
3519 | if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && |
3520 | (sbi->s_mount_state & EXT4_VALID_FS)) | |
ac27a0ec DK |
3521 | es->s_state = cpu_to_le16(sbi->s_mount_state); |
3522 | ||
a63c9eb2 | 3523 | if (sbi->s_journal) |
0390131b | 3524 | ext4_mark_recovery_complete(sb, es); |
ac27a0ec | 3525 | } else { |
a13fb1a4 ES |
3526 | /* Make sure we can mount this feature set readwrite */ |
3527 | if (!ext4_feature_set_ok(sb, 0)) { | |
ac27a0ec DK |
3528 | err = -EROFS; |
3529 | goto restore_opts; | |
3530 | } | |
8a266467 TT |
3531 | /* |
3532 | * Make sure the group descriptor checksums | |
0b8e58a1 | 3533 | * are sane. If they aren't, refuse to remount r/w. |
8a266467 TT |
3534 | */ |
3535 | for (g = 0; g < sbi->s_groups_count; g++) { | |
3536 | struct ext4_group_desc *gdp = | |
3537 | ext4_get_group_desc(sb, g, NULL); | |
3538 | ||
3539 | if (!ext4_group_desc_csum_verify(sbi, g, gdp)) { | |
b31e1552 ES |
3540 | ext4_msg(sb, KERN_ERR, |
3541 | "ext4_remount: Checksum for group %u failed (%u!=%u)", | |
8a266467 TT |
3542 | g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)), |
3543 | le16_to_cpu(gdp->bg_checksum)); | |
3544 | err = -EINVAL; | |
3545 | goto restore_opts; | |
3546 | } | |
3547 | } | |
3548 | ||
ead6596b ES |
3549 | /* |
3550 | * If we have an unprocessed orphan list hanging | |
3551 | * around from a previously readonly bdev mount, | |
3552 | * require a full umount/remount for now. | |
3553 | */ | |
3554 | if (es->s_last_orphan) { | |
b31e1552 | 3555 | ext4_msg(sb, KERN_WARNING, "Couldn't " |
ead6596b ES |
3556 | "remount RDWR because of unprocessed " |
3557 | "orphan inode list. Please " | |
b31e1552 | 3558 | "umount/remount instead"); |
ead6596b ES |
3559 | err = -EINVAL; |
3560 | goto restore_opts; | |
3561 | } | |
3562 | ||
ac27a0ec DK |
3563 | /* |
3564 | * Mounting a RDONLY partition read-write, so reread | |
3565 | * and store the current valid flag. (It may have | |
3566 | * been changed by e2fsck since we originally mounted | |
3567 | * the partition.) | |
3568 | */ | |
0390131b FM |
3569 | if (sbi->s_journal) |
3570 | ext4_clear_journal_err(sb, es); | |
ac27a0ec | 3571 | sbi->s_mount_state = le16_to_cpu(es->s_state); |
617ba13b | 3572 | if ((err = ext4_group_extend(sb, es, n_blocks_count))) |
ac27a0ec | 3573 | goto restore_opts; |
2b2d6d01 | 3574 | if (!ext4_setup_super(sb, es, 0)) |
ac27a0ec DK |
3575 | sb->s_flags &= ~MS_RDONLY; |
3576 | } | |
3577 | } | |
6fd058f7 | 3578 | ext4_setup_system_zone(sb); |
0390131b | 3579 | if (sbi->s_journal == NULL) |
e2d67052 | 3580 | ext4_commit_super(sb, 1); |
0390131b | 3581 | |
ac27a0ec DK |
3582 | #ifdef CONFIG_QUOTA |
3583 | /* Release old quota file names */ | |
3584 | for (i = 0; i < MAXQUOTAS; i++) | |
3585 | if (old_opts.s_qf_names[i] && | |
3586 | old_opts.s_qf_names[i] != sbi->s_qf_names[i]) | |
3587 | kfree(old_opts.s_qf_names[i]); | |
3588 | #endif | |
bbd6851a | 3589 | unlock_super(sb); |
337eb00a | 3590 | unlock_kernel(); |
ac27a0ec | 3591 | return 0; |
0b8e58a1 | 3592 | |
ac27a0ec DK |
3593 | restore_opts: |
3594 | sb->s_flags = old_sb_flags; | |
3595 | sbi->s_mount_opt = old_opts.s_mount_opt; | |
3596 | sbi->s_resuid = old_opts.s_resuid; | |
3597 | sbi->s_resgid = old_opts.s_resgid; | |
3598 | sbi->s_commit_interval = old_opts.s_commit_interval; | |
30773840 TT |
3599 | sbi->s_min_batch_time = old_opts.s_min_batch_time; |
3600 | sbi->s_max_batch_time = old_opts.s_max_batch_time; | |
ac27a0ec DK |
3601 | #ifdef CONFIG_QUOTA |
3602 | sbi->s_jquota_fmt = old_opts.s_jquota_fmt; | |
3603 | for (i = 0; i < MAXQUOTAS; i++) { | |
3604 | if (sbi->s_qf_names[i] && | |
3605 | old_opts.s_qf_names[i] != sbi->s_qf_names[i]) | |
3606 | kfree(sbi->s_qf_names[i]); | |
3607 | sbi->s_qf_names[i] = old_opts.s_qf_names[i]; | |
3608 | } | |
3609 | #endif | |
bbd6851a | 3610 | unlock_super(sb); |
337eb00a | 3611 | unlock_kernel(); |
ac27a0ec DK |
3612 | return err; |
3613 | } | |
3614 | ||
2b2d6d01 | 3615 | static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) |
ac27a0ec DK |
3616 | { |
3617 | struct super_block *sb = dentry->d_sb; | |
617ba13b MC |
3618 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
3619 | struct ext4_super_block *es = sbi->s_es; | |
960cc398 | 3620 | u64 fsid; |
ac27a0ec | 3621 | |
5e70030d BP |
3622 | if (test_opt(sb, MINIX_DF)) { |
3623 | sbi->s_overhead_last = 0; | |
6bc9feff | 3624 | } else if (sbi->s_blocks_last != ext4_blocks_count(es)) { |
8df9675f | 3625 | ext4_group_t i, ngroups = ext4_get_groups_count(sb); |
5e70030d | 3626 | ext4_fsblk_t overhead = 0; |
ac27a0ec DK |
3627 | |
3628 | /* | |
5e70030d BP |
3629 | * Compute the overhead (FS structures). This is constant |
3630 | * for a given filesystem unless the number of block groups | |
3631 | * changes so we cache the previous value until it does. | |
ac27a0ec DK |
3632 | */ |
3633 | ||
3634 | /* | |
3635 | * All of the blocks before first_data_block are | |
3636 | * overhead | |
3637 | */ | |
3638 | overhead = le32_to_cpu(es->s_first_data_block); | |
3639 | ||
3640 | /* | |
3641 | * Add the overhead attributed to the superblock and | |
3642 | * block group descriptors. If the sparse superblocks | |
3643 | * feature is turned on, then not all groups have this. | |
3644 | */ | |
3645 | for (i = 0; i < ngroups; i++) { | |
617ba13b MC |
3646 | overhead += ext4_bg_has_super(sb, i) + |
3647 | ext4_bg_num_gdb(sb, i); | |
ac27a0ec DK |
3648 | cond_resched(); |
3649 | } | |
3650 | ||
3651 | /* | |
3652 | * Every block group has an inode bitmap, a block | |
3653 | * bitmap, and an inode table. | |
3654 | */ | |
5e70030d BP |
3655 | overhead += ngroups * (2 + sbi->s_itb_per_group); |
3656 | sbi->s_overhead_last = overhead; | |
3657 | smp_wmb(); | |
6bc9feff | 3658 | sbi->s_blocks_last = ext4_blocks_count(es); |
ac27a0ec DK |
3659 | } |
3660 | ||
617ba13b | 3661 | buf->f_type = EXT4_SUPER_MAGIC; |
ac27a0ec | 3662 | buf->f_bsize = sb->s_blocksize; |
5e70030d | 3663 | buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last; |
6bc6e63f AK |
3664 | buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) - |
3665 | percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter); | |
308ba3ec | 3666 | ext4_free_blocks_count_set(es, buf->f_bfree); |
bd81d8ee LV |
3667 | buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es); |
3668 | if (buf->f_bfree < ext4_r_blocks_count(es)) | |
ac27a0ec DK |
3669 | buf->f_bavail = 0; |
3670 | buf->f_files = le32_to_cpu(es->s_inodes_count); | |
52d9f3b4 | 3671 | buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); |
5e70030d | 3672 | es->s_free_inodes_count = cpu_to_le32(buf->f_ffree); |
617ba13b | 3673 | buf->f_namelen = EXT4_NAME_LEN; |
960cc398 PE |
3674 | fsid = le64_to_cpup((void *)es->s_uuid) ^ |
3675 | le64_to_cpup((void *)es->s_uuid + sizeof(u64)); | |
3676 | buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; | |
3677 | buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; | |
0b8e58a1 | 3678 | |
ac27a0ec DK |
3679 | return 0; |
3680 | } | |
3681 | ||
0b8e58a1 AD |
3682 | /* Helper function for writing quotas on sync - we need to start transaction |
3683 | * before quota file is locked for write. Otherwise the are possible deadlocks: | |
ac27a0ec | 3684 | * Process 1 Process 2 |
617ba13b | 3685 | * ext4_create() quota_sync() |
a269eb18 JK |
3686 | * jbd2_journal_start() write_dquot() |
3687 | * vfs_dq_init() down(dqio_mutex) | |
dab291af | 3688 | * down(dqio_mutex) jbd2_journal_start() |
ac27a0ec DK |
3689 | * |
3690 | */ | |
3691 | ||
3692 | #ifdef CONFIG_QUOTA | |
3693 | ||
3694 | static inline struct inode *dquot_to_inode(struct dquot *dquot) | |
3695 | { | |
3696 | return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type]; | |
3697 | } | |
3698 | ||
617ba13b | 3699 | static int ext4_write_dquot(struct dquot *dquot) |
ac27a0ec DK |
3700 | { |
3701 | int ret, err; | |
3702 | handle_t *handle; | |
3703 | struct inode *inode; | |
3704 | ||
3705 | inode = dquot_to_inode(dquot); | |
617ba13b | 3706 | handle = ext4_journal_start(inode, |
0b8e58a1 | 3707 | EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); |
ac27a0ec DK |
3708 | if (IS_ERR(handle)) |
3709 | return PTR_ERR(handle); | |
3710 | ret = dquot_commit(dquot); | |
617ba13b | 3711 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
3712 | if (!ret) |
3713 | ret = err; | |
3714 | return ret; | |
3715 | } | |
3716 | ||
617ba13b | 3717 | static int ext4_acquire_dquot(struct dquot *dquot) |
ac27a0ec DK |
3718 | { |
3719 | int ret, err; | |
3720 | handle_t *handle; | |
3721 | ||
617ba13b | 3722 | handle = ext4_journal_start(dquot_to_inode(dquot), |
0b8e58a1 | 3723 | EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); |
ac27a0ec DK |
3724 | if (IS_ERR(handle)) |
3725 | return PTR_ERR(handle); | |
3726 | ret = dquot_acquire(dquot); | |
617ba13b | 3727 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
3728 | if (!ret) |
3729 | ret = err; | |
3730 | return ret; | |
3731 | } | |
3732 | ||
617ba13b | 3733 | static int ext4_release_dquot(struct dquot *dquot) |
ac27a0ec DK |
3734 | { |
3735 | int ret, err; | |
3736 | handle_t *handle; | |
3737 | ||
617ba13b | 3738 | handle = ext4_journal_start(dquot_to_inode(dquot), |
0b8e58a1 | 3739 | EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); |
9c3013e9 JK |
3740 | if (IS_ERR(handle)) { |
3741 | /* Release dquot anyway to avoid endless cycle in dqput() */ | |
3742 | dquot_release(dquot); | |
ac27a0ec | 3743 | return PTR_ERR(handle); |
9c3013e9 | 3744 | } |
ac27a0ec | 3745 | ret = dquot_release(dquot); |
617ba13b | 3746 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
3747 | if (!ret) |
3748 | ret = err; | |
3749 | return ret; | |
3750 | } | |
3751 | ||
617ba13b | 3752 | static int ext4_mark_dquot_dirty(struct dquot *dquot) |
ac27a0ec | 3753 | { |
2c8be6b2 | 3754 | /* Are we journaling quotas? */ |
617ba13b MC |
3755 | if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] || |
3756 | EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) { | |
ac27a0ec | 3757 | dquot_mark_dquot_dirty(dquot); |
617ba13b | 3758 | return ext4_write_dquot(dquot); |
ac27a0ec DK |
3759 | } else { |
3760 | return dquot_mark_dquot_dirty(dquot); | |
3761 | } | |
3762 | } | |
3763 | ||
617ba13b | 3764 | static int ext4_write_info(struct super_block *sb, int type) |
ac27a0ec DK |
3765 | { |
3766 | int ret, err; | |
3767 | handle_t *handle; | |
3768 | ||
3769 | /* Data block + inode block */ | |
617ba13b | 3770 | handle = ext4_journal_start(sb->s_root->d_inode, 2); |
ac27a0ec DK |
3771 | if (IS_ERR(handle)) |
3772 | return PTR_ERR(handle); | |
3773 | ret = dquot_commit_info(sb, type); | |
617ba13b | 3774 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
3775 | if (!ret) |
3776 | ret = err; | |
3777 | return ret; | |
3778 | } | |
3779 | ||
3780 | /* | |
3781 | * Turn on quotas during mount time - we need to find | |
3782 | * the quota file and such... | |
3783 | */ | |
617ba13b | 3784 | static int ext4_quota_on_mount(struct super_block *sb, int type) |
ac27a0ec | 3785 | { |
617ba13b | 3786 | return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type], |
0b8e58a1 | 3787 | EXT4_SB(sb)->s_jquota_fmt, type); |
ac27a0ec DK |
3788 | } |
3789 | ||
3790 | /* | |
3791 | * Standard function to be called on quota_on | |
3792 | */ | |
617ba13b | 3793 | static int ext4_quota_on(struct super_block *sb, int type, int format_id, |
8264613d | 3794 | char *name, int remount) |
ac27a0ec DK |
3795 | { |
3796 | int err; | |
8264613d | 3797 | struct path path; |
ac27a0ec DK |
3798 | |
3799 | if (!test_opt(sb, QUOTA)) | |
3800 | return -EINVAL; | |
8264613d | 3801 | /* When remounting, no checks are needed and in fact, name is NULL */ |
0623543b | 3802 | if (remount) |
8264613d | 3803 | return vfs_quota_on(sb, type, format_id, name, remount); |
0623543b | 3804 | |
8264613d | 3805 | err = kern_path(name, LOOKUP_FOLLOW, &path); |
ac27a0ec DK |
3806 | if (err) |
3807 | return err; | |
0623543b | 3808 | |
ac27a0ec | 3809 | /* Quotafile not on the same filesystem? */ |
8264613d AV |
3810 | if (path.mnt->mnt_sb != sb) { |
3811 | path_put(&path); | |
ac27a0ec DK |
3812 | return -EXDEV; |
3813 | } | |
0623543b JK |
3814 | /* Journaling quota? */ |
3815 | if (EXT4_SB(sb)->s_qf_names[type]) { | |
2b2d6d01 | 3816 | /* Quotafile not in fs root? */ |
8264613d | 3817 | if (path.dentry->d_parent != sb->s_root) |
b31e1552 ES |
3818 | ext4_msg(sb, KERN_WARNING, |
3819 | "Quota file not on filesystem root. " | |
3820 | "Journaled quota will not work"); | |
2b2d6d01 | 3821 | } |
0623543b JK |
3822 | |
3823 | /* | |
3824 | * When we journal data on quota file, we have to flush journal to see | |
3825 | * all updates to the file when we bypass pagecache... | |
3826 | */ | |
0390131b FM |
3827 | if (EXT4_SB(sb)->s_journal && |
3828 | ext4_should_journal_data(path.dentry->d_inode)) { | |
0623543b JK |
3829 | /* |
3830 | * We don't need to lock updates but journal_flush() could | |
3831 | * otherwise be livelocked... | |
3832 | */ | |
3833 | jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); | |
7ffe1ea8 | 3834 | err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); |
0623543b | 3835 | jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); |
7ffe1ea8 | 3836 | if (err) { |
8264613d | 3837 | path_put(&path); |
7ffe1ea8 HK |
3838 | return err; |
3839 | } | |
0623543b JK |
3840 | } |
3841 | ||
8264613d AV |
3842 | err = vfs_quota_on_path(sb, type, format_id, &path); |
3843 | path_put(&path); | |
77e69dac | 3844 | return err; |
ac27a0ec DK |
3845 | } |
3846 | ||
3847 | /* Read data from quotafile - avoid pagecache and such because we cannot afford | |
3848 | * acquiring the locks... As quota files are never truncated and quota code | |
3849 | * itself serializes the operations (and noone else should touch the files) | |
3850 | * we don't have to be afraid of races */ | |
617ba13b | 3851 | static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, |
ac27a0ec DK |
3852 | size_t len, loff_t off) |
3853 | { | |
3854 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
725d26d3 | 3855 | ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); |
ac27a0ec DK |
3856 | int err = 0; |
3857 | int offset = off & (sb->s_blocksize - 1); | |
3858 | int tocopy; | |
3859 | size_t toread; | |
3860 | struct buffer_head *bh; | |
3861 | loff_t i_size = i_size_read(inode); | |
3862 | ||
3863 | if (off > i_size) | |
3864 | return 0; | |
3865 | if (off+len > i_size) | |
3866 | len = i_size-off; | |
3867 | toread = len; | |
3868 | while (toread > 0) { | |
3869 | tocopy = sb->s_blocksize - offset < toread ? | |
3870 | sb->s_blocksize - offset : toread; | |
617ba13b | 3871 | bh = ext4_bread(NULL, inode, blk, 0, &err); |
ac27a0ec DK |
3872 | if (err) |
3873 | return err; | |
3874 | if (!bh) /* A hole? */ | |
3875 | memset(data, 0, tocopy); | |
3876 | else | |
3877 | memcpy(data, bh->b_data+offset, tocopy); | |
3878 | brelse(bh); | |
3879 | offset = 0; | |
3880 | toread -= tocopy; | |
3881 | data += tocopy; | |
3882 | blk++; | |
3883 | } | |
3884 | return len; | |
3885 | } | |
3886 | ||
3887 | /* Write to quotafile (we know the transaction is already started and has | |
3888 | * enough credits) */ | |
617ba13b | 3889 | static ssize_t ext4_quota_write(struct super_block *sb, int type, |
ac27a0ec DK |
3890 | const char *data, size_t len, loff_t off) |
3891 | { | |
3892 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
725d26d3 | 3893 | ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); |
ac27a0ec DK |
3894 | int err = 0; |
3895 | int offset = off & (sb->s_blocksize - 1); | |
3896 | int tocopy; | |
617ba13b | 3897 | int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL; |
ac27a0ec DK |
3898 | size_t towrite = len; |
3899 | struct buffer_head *bh; | |
3900 | handle_t *handle = journal_current_handle(); | |
3901 | ||
0390131b | 3902 | if (EXT4_SB(sb)->s_journal && !handle) { |
b31e1552 ES |
3903 | ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" |
3904 | " cancelled because transaction is not started", | |
9c3013e9 JK |
3905 | (unsigned long long)off, (unsigned long long)len); |
3906 | return -EIO; | |
3907 | } | |
ac27a0ec DK |
3908 | mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA); |
3909 | while (towrite > 0) { | |
3910 | tocopy = sb->s_blocksize - offset < towrite ? | |
3911 | sb->s_blocksize - offset : towrite; | |
617ba13b | 3912 | bh = ext4_bread(handle, inode, blk, 1, &err); |
ac27a0ec DK |
3913 | if (!bh) |
3914 | goto out; | |
3915 | if (journal_quota) { | |
617ba13b | 3916 | err = ext4_journal_get_write_access(handle, bh); |
ac27a0ec DK |
3917 | if (err) { |
3918 | brelse(bh); | |
3919 | goto out; | |
3920 | } | |
3921 | } | |
3922 | lock_buffer(bh); | |
3923 | memcpy(bh->b_data+offset, data, tocopy); | |
3924 | flush_dcache_page(bh->b_page); | |
3925 | unlock_buffer(bh); | |
3926 | if (journal_quota) | |
0390131b | 3927 | err = ext4_handle_dirty_metadata(handle, NULL, bh); |
ac27a0ec DK |
3928 | else { |
3929 | /* Always do at least ordered writes for quotas */ | |
678aaf48 | 3930 | err = ext4_jbd2_file_inode(handle, inode); |
ac27a0ec DK |
3931 | mark_buffer_dirty(bh); |
3932 | } | |
3933 | brelse(bh); | |
3934 | if (err) | |
3935 | goto out; | |
3936 | offset = 0; | |
3937 | towrite -= tocopy; | |
3938 | data += tocopy; | |
3939 | blk++; | |
3940 | } | |
3941 | out: | |
4d04e4fb JK |
3942 | if (len == towrite) { |
3943 | mutex_unlock(&inode->i_mutex); | |
ac27a0ec | 3944 | return err; |
4d04e4fb | 3945 | } |
ac27a0ec DK |
3946 | if (inode->i_size < off+len-towrite) { |
3947 | i_size_write(inode, off+len-towrite); | |
617ba13b | 3948 | EXT4_I(inode)->i_disksize = inode->i_size; |
ac27a0ec | 3949 | } |
ac27a0ec | 3950 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; |
617ba13b | 3951 | ext4_mark_inode_dirty(handle, inode); |
ac27a0ec DK |
3952 | mutex_unlock(&inode->i_mutex); |
3953 | return len - towrite; | |
3954 | } | |
3955 | ||
3956 | #endif | |
3957 | ||
0b8e58a1 AD |
3958 | static int ext4_get_sb(struct file_system_type *fs_type, int flags, |
3959 | const char *dev_name, void *data, struct vfsmount *mnt) | |
ac27a0ec | 3960 | { |
0b8e58a1 | 3961 | return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt); |
ac27a0ec DK |
3962 | } |
3963 | ||
03010a33 TT |
3964 | static struct file_system_type ext4_fs_type = { |
3965 | .owner = THIS_MODULE, | |
3966 | .name = "ext4", | |
3967 | .get_sb = ext4_get_sb, | |
3968 | .kill_sb = kill_block_super, | |
3969 | .fs_flags = FS_REQUIRES_DEV, | |
3970 | }; | |
3971 | ||
3972 | #ifdef CONFIG_EXT4DEV_COMPAT | |
0b8e58a1 AD |
3973 | static int ext4dev_get_sb(struct file_system_type *fs_type, int flags, |
3974 | const char *dev_name, void *data,struct vfsmount *mnt) | |
03010a33 | 3975 | { |
b31e1552 ES |
3976 | printk(KERN_WARNING "EXT4-fs (%s): Update your userspace programs " |
3977 | "to mount using ext4\n", dev_name); | |
3978 | printk(KERN_WARNING "EXT4-fs (%s): ext4dev backwards compatibility " | |
3979 | "will go away by 2.6.31\n", dev_name); | |
0b8e58a1 | 3980 | return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt); |
03010a33 TT |
3981 | } |
3982 | ||
617ba13b | 3983 | static struct file_system_type ext4dev_fs_type = { |
ac27a0ec | 3984 | .owner = THIS_MODULE, |
617ba13b | 3985 | .name = "ext4dev", |
03010a33 | 3986 | .get_sb = ext4dev_get_sb, |
ac27a0ec DK |
3987 | .kill_sb = kill_block_super, |
3988 | .fs_flags = FS_REQUIRES_DEV, | |
3989 | }; | |
03010a33 TT |
3990 | MODULE_ALIAS("ext4dev"); |
3991 | #endif | |
ac27a0ec | 3992 | |
617ba13b | 3993 | static int __init init_ext4_fs(void) |
ac27a0ec | 3994 | { |
c9de560d AT |
3995 | int err; |
3996 | ||
6fd058f7 TT |
3997 | err = init_ext4_system_zone(); |
3998 | if (err) | |
3999 | return err; | |
3197ebdb TT |
4000 | ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj); |
4001 | if (!ext4_kset) | |
6fd058f7 | 4002 | goto out4; |
9f6200bb | 4003 | ext4_proc_root = proc_mkdir("fs/ext4", NULL); |
c9de560d | 4004 | err = init_ext4_mballoc(); |
ac27a0ec | 4005 | if (err) |
6fd058f7 | 4006 | goto out3; |
c9de560d AT |
4007 | |
4008 | err = init_ext4_xattr(); | |
4009 | if (err) | |
4010 | goto out2; | |
ac27a0ec DK |
4011 | err = init_inodecache(); |
4012 | if (err) | |
4013 | goto out1; | |
03010a33 | 4014 | err = register_filesystem(&ext4_fs_type); |
ac27a0ec DK |
4015 | if (err) |
4016 | goto out; | |
03010a33 TT |
4017 | #ifdef CONFIG_EXT4DEV_COMPAT |
4018 | err = register_filesystem(&ext4dev_fs_type); | |
4019 | if (err) { | |
4020 | unregister_filesystem(&ext4_fs_type); | |
4021 | goto out; | |
4022 | } | |
4023 | #endif | |
ac27a0ec DK |
4024 | return 0; |
4025 | out: | |
4026 | destroy_inodecache(); | |
4027 | out1: | |
617ba13b | 4028 | exit_ext4_xattr(); |
c9de560d AT |
4029 | out2: |
4030 | exit_ext4_mballoc(); | |
6fd058f7 TT |
4031 | out3: |
4032 | remove_proc_entry("fs/ext4", NULL); | |
4033 | kset_unregister(ext4_kset); | |
4034 | out4: | |
4035 | exit_ext4_system_zone(); | |
ac27a0ec DK |
4036 | return err; |
4037 | } | |
4038 | ||
617ba13b | 4039 | static void __exit exit_ext4_fs(void) |
ac27a0ec | 4040 | { |
03010a33 TT |
4041 | unregister_filesystem(&ext4_fs_type); |
4042 | #ifdef CONFIG_EXT4DEV_COMPAT | |
617ba13b | 4043 | unregister_filesystem(&ext4dev_fs_type); |
03010a33 | 4044 | #endif |
ac27a0ec | 4045 | destroy_inodecache(); |
617ba13b | 4046 | exit_ext4_xattr(); |
c9de560d | 4047 | exit_ext4_mballoc(); |
9f6200bb | 4048 | remove_proc_entry("fs/ext4", NULL); |
3197ebdb | 4049 | kset_unregister(ext4_kset); |
6fd058f7 | 4050 | exit_ext4_system_zone(); |
ac27a0ec DK |
4051 | } |
4052 | ||
4053 | MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); | |
83982b6f | 4054 | MODULE_DESCRIPTION("Fourth Extended Filesystem"); |
ac27a0ec | 4055 | MODULE_LICENSE("GPL"); |
617ba13b MC |
4056 | module_init(init_ext4_fs) |
4057 | module_exit(exit_ext4_fs) |