]>
Commit | Line | Data |
---|---|---|
ac27a0ec | 1 | /* |
617ba13b | 2 | * linux/fs/ext4/super.c |
ac27a0ec DK |
3 | * |
4 | * Copyright (C) 1992, 1993, 1994, 1995 | |
5 | * Remy Card (card@masi.ibp.fr) | |
6 | * Laboratoire MASI - Institut Blaise Pascal | |
7 | * Universite Pierre et Marie Curie (Paris VI) | |
8 | * | |
9 | * from | |
10 | * | |
11 | * linux/fs/minix/inode.c | |
12 | * | |
13 | * Copyright (C) 1991, 1992 Linus Torvalds | |
14 | * | |
15 | * Big-endian to little-endian byte-swapping/bitmaps by | |
16 | * David S. Miller (davem@caip.rutgers.edu), 1995 | |
17 | */ | |
18 | ||
19 | #include <linux/module.h> | |
20 | #include <linux/string.h> | |
21 | #include <linux/fs.h> | |
22 | #include <linux/time.h> | |
c5ca7c76 | 23 | #include <linux/vmalloc.h> |
dab291af | 24 | #include <linux/jbd2.h> |
ac27a0ec DK |
25 | #include <linux/slab.h> |
26 | #include <linux/init.h> | |
27 | #include <linux/blkdev.h> | |
28 | #include <linux/parser.h> | |
ac27a0ec | 29 | #include <linux/buffer_head.h> |
a5694255 | 30 | #include <linux/exportfs.h> |
ac27a0ec DK |
31 | #include <linux/vfs.h> |
32 | #include <linux/random.h> | |
33 | #include <linux/mount.h> | |
34 | #include <linux/namei.h> | |
35 | #include <linux/quotaops.h> | |
36 | #include <linux/seq_file.h> | |
9f6200bb | 37 | #include <linux/proc_fs.h> |
3197ebdb | 38 | #include <linux/ctype.h> |
1330593e | 39 | #include <linux/log2.h> |
717d50e4 | 40 | #include <linux/crc16.h> |
ac27a0ec DK |
41 | #include <asm/uaccess.h> |
42 | ||
bfff6873 LC |
43 | #include <linux/kthread.h> |
44 | #include <linux/freezer.h> | |
45 | ||
3dcf5451 CH |
46 | #include "ext4.h" |
47 | #include "ext4_jbd2.h" | |
ac27a0ec DK |
48 | #include "xattr.h" |
49 | #include "acl.h" | |
3661d286 | 50 | #include "mballoc.h" |
ac27a0ec | 51 | |
9bffad1e TT |
52 | #define CREATE_TRACE_POINTS |
53 | #include <trace/events/ext4.h> | |
54 | ||
1f109d5a | 55 | static struct proc_dir_entry *ext4_proc_root; |
3197ebdb | 56 | static struct kset *ext4_kset; |
0b75a840 LC |
57 | static struct ext4_lazy_init *ext4_li_info; |
58 | static struct mutex ext4_li_mtx; | |
59 | static struct ext4_features *ext4_feat; | |
9f6200bb | 60 | |
617ba13b | 61 | static int ext4_load_journal(struct super_block *, struct ext4_super_block *, |
ac27a0ec | 62 | unsigned long journal_devnum); |
e2d67052 | 63 | static int ext4_commit_super(struct super_block *sb, int sync); |
2b2d6d01 TT |
64 | static void ext4_mark_recovery_complete(struct super_block *sb, |
65 | struct ext4_super_block *es); | |
66 | static void ext4_clear_journal_err(struct super_block *sb, | |
67 | struct ext4_super_block *es); | |
617ba13b | 68 | static int ext4_sync_fs(struct super_block *sb, int wait); |
2b2d6d01 | 69 | static const char *ext4_decode_error(struct super_block *sb, int errno, |
ac27a0ec | 70 | char nbuf[16]); |
2b2d6d01 TT |
71 | static int ext4_remount(struct super_block *sb, int *flags, char *data); |
72 | static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); | |
c4be0c1d | 73 | static int ext4_unfreeze(struct super_block *sb); |
2b2d6d01 | 74 | static void ext4_write_super(struct super_block *sb); |
c4be0c1d | 75 | static int ext4_freeze(struct super_block *sb); |
152a0836 AV |
76 | static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, |
77 | const char *dev_name, void *data); | |
2035e776 TT |
78 | static inline int ext2_feature_set_ok(struct super_block *sb); |
79 | static inline int ext3_feature_set_ok(struct super_block *sb); | |
d39195c3 | 80 | static int ext4_feature_set_ok(struct super_block *sb, int readonly); |
bfff6873 LC |
81 | static void ext4_destroy_lazyinit_thread(void); |
82 | static void ext4_unregister_li_request(struct super_block *sb); | |
8f1f7453 | 83 | static void ext4_clear_request_list(void); |
ac27a0ec | 84 | |
2035e776 TT |
85 | #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) |
86 | static struct file_system_type ext2_fs_type = { | |
87 | .owner = THIS_MODULE, | |
88 | .name = "ext2", | |
89 | .mount = ext4_mount, | |
90 | .kill_sb = kill_block_super, | |
91 | .fs_flags = FS_REQUIRES_DEV, | |
92 | }; | |
93 | #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) | |
94 | #else | |
95 | #define IS_EXT2_SB(sb) (0) | |
96 | #endif | |
97 | ||
98 | ||
ba69f9ab JK |
99 | #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) |
100 | static struct file_system_type ext3_fs_type = { | |
101 | .owner = THIS_MODULE, | |
102 | .name = "ext3", | |
152a0836 | 103 | .mount = ext4_mount, |
ba69f9ab JK |
104 | .kill_sb = kill_block_super, |
105 | .fs_flags = FS_REQUIRES_DEV, | |
106 | }; | |
107 | #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) | |
108 | #else | |
109 | #define IS_EXT3_SB(sb) (0) | |
110 | #endif | |
bd81d8ee | 111 | |
8fadc143 AR |
112 | ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, |
113 | struct ext4_group_desc *bg) | |
bd81d8ee | 114 | { |
3a14589c | 115 | return le32_to_cpu(bg->bg_block_bitmap_lo) | |
8fadc143 | 116 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 117 | (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); |
bd81d8ee LV |
118 | } |
119 | ||
8fadc143 AR |
120 | ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, |
121 | struct ext4_group_desc *bg) | |
bd81d8ee | 122 | { |
5272f837 | 123 | return le32_to_cpu(bg->bg_inode_bitmap_lo) | |
8fadc143 | 124 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 125 | (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); |
bd81d8ee LV |
126 | } |
127 | ||
8fadc143 AR |
128 | ext4_fsblk_t ext4_inode_table(struct super_block *sb, |
129 | struct ext4_group_desc *bg) | |
bd81d8ee | 130 | { |
5272f837 | 131 | return le32_to_cpu(bg->bg_inode_table_lo) | |
8fadc143 | 132 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 133 | (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); |
bd81d8ee LV |
134 | } |
135 | ||
560671a0 AK |
136 | __u32 ext4_free_blks_count(struct super_block *sb, |
137 | struct ext4_group_desc *bg) | |
138 | { | |
139 | return le16_to_cpu(bg->bg_free_blocks_count_lo) | | |
140 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 141 | (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); |
560671a0 AK |
142 | } |
143 | ||
144 | __u32 ext4_free_inodes_count(struct super_block *sb, | |
145 | struct ext4_group_desc *bg) | |
146 | { | |
147 | return le16_to_cpu(bg->bg_free_inodes_count_lo) | | |
148 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 149 | (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); |
560671a0 AK |
150 | } |
151 | ||
152 | __u32 ext4_used_dirs_count(struct super_block *sb, | |
153 | struct ext4_group_desc *bg) | |
154 | { | |
155 | return le16_to_cpu(bg->bg_used_dirs_count_lo) | | |
156 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 157 | (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); |
560671a0 AK |
158 | } |
159 | ||
160 | __u32 ext4_itable_unused_count(struct super_block *sb, | |
161 | struct ext4_group_desc *bg) | |
162 | { | |
163 | return le16_to_cpu(bg->bg_itable_unused_lo) | | |
164 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 165 | (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); |
560671a0 AK |
166 | } |
167 | ||
8fadc143 AR |
168 | void ext4_block_bitmap_set(struct super_block *sb, |
169 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 170 | { |
3a14589c | 171 | bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
172 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
173 | bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
174 | } |
175 | ||
8fadc143 AR |
176 | void ext4_inode_bitmap_set(struct super_block *sb, |
177 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 178 | { |
5272f837 | 179 | bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
180 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
181 | bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
182 | } |
183 | ||
8fadc143 AR |
184 | void ext4_inode_table_set(struct super_block *sb, |
185 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 186 | { |
5272f837 | 187 | bg->bg_inode_table_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
188 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
189 | bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
190 | } |
191 | ||
560671a0 AK |
192 | void ext4_free_blks_set(struct super_block *sb, |
193 | struct ext4_group_desc *bg, __u32 count) | |
194 | { | |
195 | bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); | |
196 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
197 | bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); | |
198 | } | |
199 | ||
200 | void ext4_free_inodes_set(struct super_block *sb, | |
201 | struct ext4_group_desc *bg, __u32 count) | |
202 | { | |
203 | bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); | |
204 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
205 | bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); | |
206 | } | |
207 | ||
208 | void ext4_used_dirs_set(struct super_block *sb, | |
209 | struct ext4_group_desc *bg, __u32 count) | |
210 | { | |
211 | bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); | |
212 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
213 | bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); | |
214 | } | |
215 | ||
216 | void ext4_itable_unused_set(struct super_block *sb, | |
217 | struct ext4_group_desc *bg, __u32 count) | |
218 | { | |
219 | bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); | |
220 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
221 | bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); | |
222 | } | |
223 | ||
d3d1faf6 CW |
224 | |
225 | /* Just increment the non-pointer handle value */ | |
226 | static handle_t *ext4_get_nojournal(void) | |
227 | { | |
228 | handle_t *handle = current->journal_info; | |
229 | unsigned long ref_cnt = (unsigned long)handle; | |
230 | ||
231 | BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT); | |
232 | ||
233 | ref_cnt++; | |
234 | handle = (handle_t *)ref_cnt; | |
235 | ||
236 | current->journal_info = handle; | |
237 | return handle; | |
238 | } | |
239 | ||
240 | ||
241 | /* Decrement the non-pointer handle value */ | |
242 | static void ext4_put_nojournal(handle_t *handle) | |
243 | { | |
244 | unsigned long ref_cnt = (unsigned long)handle; | |
245 | ||
246 | BUG_ON(ref_cnt == 0); | |
247 | ||
248 | ref_cnt--; | |
249 | handle = (handle_t *)ref_cnt; | |
250 | ||
251 | current->journal_info = handle; | |
252 | } | |
253 | ||
ac27a0ec | 254 | /* |
dab291af | 255 | * Wrappers for jbd2_journal_start/end. |
ac27a0ec DK |
256 | * |
257 | * The only special thing we need to do here is to make sure that all | |
258 | * journal_end calls result in the superblock being marked dirty, so | |
259 | * that sync() will call the filesystem's write_super callback if | |
260 | * appropriate. | |
be4f27d3 YY |
261 | * |
262 | * To avoid j_barrier hold in userspace when a user calls freeze(), | |
263 | * ext4 prevents a new handle from being started by s_frozen, which | |
264 | * is in an upper layer. | |
ac27a0ec | 265 | */ |
617ba13b | 266 | handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks) |
ac27a0ec DK |
267 | { |
268 | journal_t *journal; | |
be4f27d3 | 269 | handle_t *handle; |
ac27a0ec DK |
270 | |
271 | if (sb->s_flags & MS_RDONLY) | |
272 | return ERR_PTR(-EROFS); | |
273 | ||
617ba13b | 274 | journal = EXT4_SB(sb)->s_journal; |
be4f27d3 YY |
275 | handle = ext4_journal_current_handle(); |
276 | ||
277 | /* | |
278 | * If a handle has been started, it should be allowed to | |
279 | * finish, otherwise deadlock could happen between freeze | |
280 | * and others(e.g. truncate) due to the restart of the | |
281 | * journal handle if the filesystem is forzen and active | |
282 | * handles are not stopped. | |
283 | */ | |
284 | if (!handle) | |
285 | vfs_check_frozen(sb, SB_FREEZE_TRANS); | |
286 | ||
287 | if (!journal) | |
288 | return ext4_get_nojournal(); | |
289 | /* | |
290 | * Special case here: if the journal has aborted behind our | |
291 | * backs (eg. EIO in the commit thread), then we still need to | |
292 | * take the FS itself readonly cleanly. | |
293 | */ | |
294 | if (is_journal_aborted(journal)) { | |
295 | ext4_abort(sb, "Detected aborted journal"); | |
296 | return ERR_PTR(-EROFS); | |
ac27a0ec | 297 | } |
be4f27d3 | 298 | return jbd2_journal_start(journal, nblocks); |
ac27a0ec DK |
299 | } |
300 | ||
301 | /* | |
302 | * The only special thing we need to do here is to make sure that all | |
dab291af | 303 | * jbd2_journal_stop calls result in the superblock being marked dirty, so |
ac27a0ec DK |
304 | * that sync() will call the filesystem's write_super callback if |
305 | * appropriate. | |
306 | */ | |
c398eda0 | 307 | int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle) |
ac27a0ec DK |
308 | { |
309 | struct super_block *sb; | |
310 | int err; | |
311 | int rc; | |
312 | ||
0390131b | 313 | if (!ext4_handle_valid(handle)) { |
d3d1faf6 | 314 | ext4_put_nojournal(handle); |
0390131b FM |
315 | return 0; |
316 | } | |
ac27a0ec DK |
317 | sb = handle->h_transaction->t_journal->j_private; |
318 | err = handle->h_err; | |
dab291af | 319 | rc = jbd2_journal_stop(handle); |
ac27a0ec DK |
320 | |
321 | if (!err) | |
322 | err = rc; | |
323 | if (err) | |
c398eda0 | 324 | __ext4_std_error(sb, where, line, err); |
ac27a0ec DK |
325 | return err; |
326 | } | |
327 | ||
90c7201b TT |
328 | void ext4_journal_abort_handle(const char *caller, unsigned int line, |
329 | const char *err_fn, struct buffer_head *bh, | |
330 | handle_t *handle, int err) | |
ac27a0ec DK |
331 | { |
332 | char nbuf[16]; | |
617ba13b | 333 | const char *errstr = ext4_decode_error(NULL, err, nbuf); |
ac27a0ec | 334 | |
0390131b FM |
335 | BUG_ON(!ext4_handle_valid(handle)); |
336 | ||
ac27a0ec DK |
337 | if (bh) |
338 | BUFFER_TRACE(bh, "abort"); | |
339 | ||
340 | if (!handle->h_err) | |
341 | handle->h_err = err; | |
342 | ||
343 | if (is_handle_aborted(handle)) | |
344 | return; | |
345 | ||
90c7201b TT |
346 | printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n", |
347 | caller, line, errstr, err_fn); | |
ac27a0ec | 348 | |
dab291af | 349 | jbd2_journal_abort_handle(handle); |
ac27a0ec DK |
350 | } |
351 | ||
1c13d5c0 TT |
352 | static void __save_error_info(struct super_block *sb, const char *func, |
353 | unsigned int line) | |
354 | { | |
355 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; | |
356 | ||
357 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; | |
358 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
359 | es->s_last_error_time = cpu_to_le32(get_seconds()); | |
360 | strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func)); | |
361 | es->s_last_error_line = cpu_to_le32(line); | |
362 | if (!es->s_first_error_time) { | |
363 | es->s_first_error_time = es->s_last_error_time; | |
364 | strncpy(es->s_first_error_func, func, | |
365 | sizeof(es->s_first_error_func)); | |
366 | es->s_first_error_line = cpu_to_le32(line); | |
367 | es->s_first_error_ino = es->s_last_error_ino; | |
368 | es->s_first_error_block = es->s_last_error_block; | |
369 | } | |
66e61a9e TT |
370 | /* |
371 | * Start the daily error reporting function if it hasn't been | |
372 | * started already | |
373 | */ | |
374 | if (!es->s_error_count) | |
375 | mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ); | |
1c13d5c0 TT |
376 | es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1); |
377 | } | |
378 | ||
379 | static void save_error_info(struct super_block *sb, const char *func, | |
380 | unsigned int line) | |
381 | { | |
382 | __save_error_info(sb, func, line); | |
383 | ext4_commit_super(sb, 1); | |
384 | } | |
385 | ||
386 | ||
ac27a0ec DK |
387 | /* Deal with the reporting of failure conditions on a filesystem such as |
388 | * inconsistencies detected or read IO failures. | |
389 | * | |
390 | * On ext2, we can store the error state of the filesystem in the | |
617ba13b | 391 | * superblock. That is not possible on ext4, because we may have other |
ac27a0ec DK |
392 | * write ordering constraints on the superblock which prevent us from |
393 | * writing it out straight away; and given that the journal is about to | |
394 | * be aborted, we can't rely on the current, or future, transactions to | |
395 | * write out the superblock safely. | |
396 | * | |
dab291af | 397 | * We'll just use the jbd2_journal_abort() error code to record an error in |
d6b198bc | 398 | * the journal instead. On recovery, the journal will complain about |
ac27a0ec DK |
399 | * that error until we've noted it down and cleared it. |
400 | */ | |
401 | ||
617ba13b | 402 | static void ext4_handle_error(struct super_block *sb) |
ac27a0ec | 403 | { |
ac27a0ec DK |
404 | if (sb->s_flags & MS_RDONLY) |
405 | return; | |
406 | ||
2b2d6d01 | 407 | if (!test_opt(sb, ERRORS_CONT)) { |
617ba13b | 408 | journal_t *journal = EXT4_SB(sb)->s_journal; |
ac27a0ec | 409 | |
4ab2f15b | 410 | EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; |
ac27a0ec | 411 | if (journal) |
dab291af | 412 | jbd2_journal_abort(journal, -EIO); |
ac27a0ec | 413 | } |
2b2d6d01 | 414 | if (test_opt(sb, ERRORS_RO)) { |
b31e1552 | 415 | ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); |
ac27a0ec DK |
416 | sb->s_flags |= MS_RDONLY; |
417 | } | |
ac27a0ec | 418 | if (test_opt(sb, ERRORS_PANIC)) |
617ba13b | 419 | panic("EXT4-fs (device %s): panic forced after error\n", |
ac27a0ec DK |
420 | sb->s_id); |
421 | } | |
422 | ||
12062ddd | 423 | void __ext4_error(struct super_block *sb, const char *function, |
c398eda0 | 424 | unsigned int line, const char *fmt, ...) |
ac27a0ec | 425 | { |
0ff2ea7d | 426 | struct va_format vaf; |
ac27a0ec DK |
427 | va_list args; |
428 | ||
429 | va_start(args, fmt); | |
0ff2ea7d JP |
430 | vaf.fmt = fmt; |
431 | vaf.va = &args; | |
432 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", | |
433 | sb->s_id, function, line, current->comm, &vaf); | |
ac27a0ec DK |
434 | va_end(args); |
435 | ||
617ba13b | 436 | ext4_handle_error(sb); |
ac27a0ec DK |
437 | } |
438 | ||
c398eda0 TT |
439 | void ext4_error_inode(struct inode *inode, const char *function, |
440 | unsigned int line, ext4_fsblk_t block, | |
273df556 FM |
441 | const char *fmt, ...) |
442 | { | |
443 | va_list args; | |
f7c21177 | 444 | struct va_format vaf; |
1c13d5c0 | 445 | struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es; |
273df556 | 446 | |
1c13d5c0 TT |
447 | es->s_last_error_ino = cpu_to_le32(inode->i_ino); |
448 | es->s_last_error_block = cpu_to_le64(block); | |
449 | save_error_info(inode->i_sb, function, line); | |
273df556 | 450 | va_start(args, fmt); |
f7c21177 TT |
451 | vaf.fmt = fmt; |
452 | vaf.va = &args; | |
c398eda0 TT |
453 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ", |
454 | inode->i_sb->s_id, function, line, inode->i_ino); | |
455 | if (block) | |
f7c21177 TT |
456 | printk(KERN_CONT "block %llu: ", block); |
457 | printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf); | |
273df556 FM |
458 | va_end(args); |
459 | ||
460 | ext4_handle_error(inode->i_sb); | |
461 | } | |
462 | ||
c398eda0 | 463 | void ext4_error_file(struct file *file, const char *function, |
f7c21177 TT |
464 | unsigned int line, ext4_fsblk_t block, |
465 | const char *fmt, ...) | |
273df556 FM |
466 | { |
467 | va_list args; | |
f7c21177 | 468 | struct va_format vaf; |
1c13d5c0 | 469 | struct ext4_super_block *es; |
273df556 FM |
470 | struct inode *inode = file->f_dentry->d_inode; |
471 | char pathname[80], *path; | |
472 | ||
1c13d5c0 TT |
473 | es = EXT4_SB(inode->i_sb)->s_es; |
474 | es->s_last_error_ino = cpu_to_le32(inode->i_ino); | |
475 | save_error_info(inode->i_sb, function, line); | |
273df556 | 476 | path = d_path(&(file->f_path), pathname, sizeof(pathname)); |
f9a62d09 | 477 | if (IS_ERR(path)) |
273df556 FM |
478 | path = "(unknown)"; |
479 | printk(KERN_CRIT | |
f7c21177 TT |
480 | "EXT4-fs error (device %s): %s:%d: inode #%lu: ", |
481 | inode->i_sb->s_id, function, line, inode->i_ino); | |
482 | if (block) | |
483 | printk(KERN_CONT "block %llu: ", block); | |
484 | va_start(args, fmt); | |
485 | vaf.fmt = fmt; | |
486 | vaf.va = &args; | |
487 | printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf); | |
273df556 FM |
488 | va_end(args); |
489 | ||
490 | ext4_handle_error(inode->i_sb); | |
491 | } | |
492 | ||
2b2d6d01 | 493 | static const char *ext4_decode_error(struct super_block *sb, int errno, |
ac27a0ec DK |
494 | char nbuf[16]) |
495 | { | |
496 | char *errstr = NULL; | |
497 | ||
498 | switch (errno) { | |
499 | case -EIO: | |
500 | errstr = "IO failure"; | |
501 | break; | |
502 | case -ENOMEM: | |
503 | errstr = "Out of memory"; | |
504 | break; | |
505 | case -EROFS: | |
78f1ddbb TT |
506 | if (!sb || (EXT4_SB(sb)->s_journal && |
507 | EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) | |
ac27a0ec DK |
508 | errstr = "Journal has aborted"; |
509 | else | |
510 | errstr = "Readonly filesystem"; | |
511 | break; | |
512 | default: | |
513 | /* If the caller passed in an extra buffer for unknown | |
514 | * errors, textualise them now. Else we just return | |
515 | * NULL. */ | |
516 | if (nbuf) { | |
517 | /* Check for truncated error codes... */ | |
518 | if (snprintf(nbuf, 16, "error %d", -errno) >= 0) | |
519 | errstr = nbuf; | |
520 | } | |
521 | break; | |
522 | } | |
523 | ||
524 | return errstr; | |
525 | } | |
526 | ||
617ba13b | 527 | /* __ext4_std_error decodes expected errors from journaling functions |
ac27a0ec DK |
528 | * automatically and invokes the appropriate error response. */ |
529 | ||
c398eda0 TT |
530 | void __ext4_std_error(struct super_block *sb, const char *function, |
531 | unsigned int line, int errno) | |
ac27a0ec DK |
532 | { |
533 | char nbuf[16]; | |
534 | const char *errstr; | |
535 | ||
536 | /* Special case: if the error is EROFS, and we're not already | |
537 | * inside a transaction, then there's really no point in logging | |
538 | * an error. */ | |
539 | if (errno == -EROFS && journal_current_handle() == NULL && | |
540 | (sb->s_flags & MS_RDONLY)) | |
541 | return; | |
542 | ||
617ba13b | 543 | errstr = ext4_decode_error(sb, errno, nbuf); |
c398eda0 TT |
544 | printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", |
545 | sb->s_id, function, line, errstr); | |
1c13d5c0 | 546 | save_error_info(sb, function, line); |
ac27a0ec | 547 | |
617ba13b | 548 | ext4_handle_error(sb); |
ac27a0ec DK |
549 | } |
550 | ||
551 | /* | |
617ba13b | 552 | * ext4_abort is a much stronger failure handler than ext4_error. The |
ac27a0ec DK |
553 | * abort function may be used to deal with unrecoverable failures such |
554 | * as journal IO errors or ENOMEM at a critical moment in log management. | |
555 | * | |
556 | * We unconditionally force the filesystem into an ABORT|READONLY state, | |
557 | * unless the error response on the fs has been set to panic in which | |
558 | * case we take the easy way out and panic immediately. | |
559 | */ | |
560 | ||
c67d859e | 561 | void __ext4_abort(struct super_block *sb, const char *function, |
c398eda0 | 562 | unsigned int line, const char *fmt, ...) |
ac27a0ec DK |
563 | { |
564 | va_list args; | |
565 | ||
1c13d5c0 | 566 | save_error_info(sb, function, line); |
ac27a0ec | 567 | va_start(args, fmt); |
c398eda0 TT |
568 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id, |
569 | function, line); | |
ac27a0ec DK |
570 | vprintk(fmt, args); |
571 | printk("\n"); | |
572 | va_end(args); | |
573 | ||
1c13d5c0 TT |
574 | if ((sb->s_flags & MS_RDONLY) == 0) { |
575 | ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); | |
576 | sb->s_flags |= MS_RDONLY; | |
577 | EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; | |
578 | if (EXT4_SB(sb)->s_journal) | |
579 | jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); | |
580 | save_error_info(sb, function, line); | |
581 | } | |
ac27a0ec | 582 | if (test_opt(sb, ERRORS_PANIC)) |
617ba13b | 583 | panic("EXT4-fs panic from previous error\n"); |
ac27a0ec DK |
584 | } |
585 | ||
0ff2ea7d | 586 | void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...) |
b31e1552 | 587 | { |
0ff2ea7d | 588 | struct va_format vaf; |
b31e1552 ES |
589 | va_list args; |
590 | ||
591 | va_start(args, fmt); | |
0ff2ea7d JP |
592 | vaf.fmt = fmt; |
593 | vaf.va = &args; | |
594 | printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); | |
b31e1552 ES |
595 | va_end(args); |
596 | } | |
597 | ||
12062ddd | 598 | void __ext4_warning(struct super_block *sb, const char *function, |
c398eda0 | 599 | unsigned int line, const char *fmt, ...) |
ac27a0ec | 600 | { |
0ff2ea7d | 601 | struct va_format vaf; |
ac27a0ec DK |
602 | va_list args; |
603 | ||
604 | va_start(args, fmt); | |
0ff2ea7d JP |
605 | vaf.fmt = fmt; |
606 | vaf.va = &args; | |
607 | printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", | |
608 | sb->s_id, function, line, &vaf); | |
ac27a0ec DK |
609 | va_end(args); |
610 | } | |
611 | ||
e29136f8 TT |
612 | void __ext4_grp_locked_error(const char *function, unsigned int line, |
613 | struct super_block *sb, ext4_group_t grp, | |
614 | unsigned long ino, ext4_fsblk_t block, | |
615 | const char *fmt, ...) | |
5d1b1b3f AK |
616 | __releases(bitlock) |
617 | __acquires(bitlock) | |
618 | { | |
0ff2ea7d | 619 | struct va_format vaf; |
5d1b1b3f AK |
620 | va_list args; |
621 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; | |
622 | ||
1c13d5c0 TT |
623 | es->s_last_error_ino = cpu_to_le32(ino); |
624 | es->s_last_error_block = cpu_to_le64(block); | |
625 | __save_error_info(sb, function, line); | |
0ff2ea7d | 626 | |
5d1b1b3f | 627 | va_start(args, fmt); |
0ff2ea7d JP |
628 | |
629 | vaf.fmt = fmt; | |
630 | vaf.va = &args; | |
21149d61 | 631 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", |
e29136f8 TT |
632 | sb->s_id, function, line, grp); |
633 | if (ino) | |
0ff2ea7d | 634 | printk(KERN_CONT "inode %lu: ", ino); |
e29136f8 | 635 | if (block) |
0ff2ea7d JP |
636 | printk(KERN_CONT "block %llu:", (unsigned long long) block); |
637 | printk(KERN_CONT "%pV\n", &vaf); | |
5d1b1b3f AK |
638 | va_end(args); |
639 | ||
640 | if (test_opt(sb, ERRORS_CONT)) { | |
e2d67052 | 641 | ext4_commit_super(sb, 0); |
5d1b1b3f AK |
642 | return; |
643 | } | |
1c13d5c0 | 644 | |
5d1b1b3f AK |
645 | ext4_unlock_group(sb, grp); |
646 | ext4_handle_error(sb); | |
647 | /* | |
648 | * We only get here in the ERRORS_RO case; relocking the group | |
649 | * may be dangerous, but nothing bad will happen since the | |
650 | * filesystem will have already been marked read/only and the | |
651 | * journal has been aborted. We return 1 as a hint to callers | |
652 | * who might what to use the return value from | |
25985edc | 653 | * ext4_grp_locked_error() to distinguish between the |
5d1b1b3f AK |
654 | * ERRORS_CONT and ERRORS_RO case, and perhaps return more |
655 | * aggressively from the ext4 function in question, with a | |
656 | * more appropriate error code. | |
657 | */ | |
658 | ext4_lock_group(sb, grp); | |
659 | return; | |
660 | } | |
661 | ||
617ba13b | 662 | void ext4_update_dynamic_rev(struct super_block *sb) |
ac27a0ec | 663 | { |
617ba13b | 664 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
ac27a0ec | 665 | |
617ba13b | 666 | if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) |
ac27a0ec DK |
667 | return; |
668 | ||
12062ddd | 669 | ext4_warning(sb, |
ac27a0ec DK |
670 | "updating to rev %d because of new feature flag, " |
671 | "running e2fsck is recommended", | |
617ba13b | 672 | EXT4_DYNAMIC_REV); |
ac27a0ec | 673 | |
617ba13b MC |
674 | es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); |
675 | es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); | |
676 | es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); | |
ac27a0ec DK |
677 | /* leave es->s_feature_*compat flags alone */ |
678 | /* es->s_uuid will be set by e2fsck if empty */ | |
679 | ||
680 | /* | |
681 | * The rest of the superblock fields should be zero, and if not it | |
682 | * means they are likely already in use, so leave them alone. We | |
683 | * can leave it up to e2fsck to clean up any inconsistencies there. | |
684 | */ | |
685 | } | |
686 | ||
687 | /* | |
688 | * Open the external journal device | |
689 | */ | |
b31e1552 | 690 | static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) |
ac27a0ec DK |
691 | { |
692 | struct block_device *bdev; | |
693 | char b[BDEVNAME_SIZE]; | |
694 | ||
d4d77629 | 695 | bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); |
ac27a0ec DK |
696 | if (IS_ERR(bdev)) |
697 | goto fail; | |
698 | return bdev; | |
699 | ||
700 | fail: | |
b31e1552 | 701 | ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld", |
ac27a0ec DK |
702 | __bdevname(dev, b), PTR_ERR(bdev)); |
703 | return NULL; | |
704 | } | |
705 | ||
706 | /* | |
707 | * Release the journal device | |
708 | */ | |
617ba13b | 709 | static int ext4_blkdev_put(struct block_device *bdev) |
ac27a0ec | 710 | { |
e525fd89 | 711 | return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); |
ac27a0ec DK |
712 | } |
713 | ||
617ba13b | 714 | static int ext4_blkdev_remove(struct ext4_sb_info *sbi) |
ac27a0ec DK |
715 | { |
716 | struct block_device *bdev; | |
717 | int ret = -ENODEV; | |
718 | ||
719 | bdev = sbi->journal_bdev; | |
720 | if (bdev) { | |
617ba13b | 721 | ret = ext4_blkdev_put(bdev); |
ac27a0ec DK |
722 | sbi->journal_bdev = NULL; |
723 | } | |
724 | return ret; | |
725 | } | |
726 | ||
727 | static inline struct inode *orphan_list_entry(struct list_head *l) | |
728 | { | |
617ba13b | 729 | return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; |
ac27a0ec DK |
730 | } |
731 | ||
617ba13b | 732 | static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) |
ac27a0ec DK |
733 | { |
734 | struct list_head *l; | |
735 | ||
b31e1552 ES |
736 | ext4_msg(sb, KERN_ERR, "sb orphan head is %d", |
737 | le32_to_cpu(sbi->s_es->s_last_orphan)); | |
ac27a0ec DK |
738 | |
739 | printk(KERN_ERR "sb_info orphan list:\n"); | |
740 | list_for_each(l, &sbi->s_orphan) { | |
741 | struct inode *inode = orphan_list_entry(l); | |
742 | printk(KERN_ERR " " | |
743 | "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", | |
744 | inode->i_sb->s_id, inode->i_ino, inode, | |
745 | inode->i_mode, inode->i_nlink, | |
746 | NEXT_ORPHAN(inode)); | |
747 | } | |
748 | } | |
749 | ||
2b2d6d01 | 750 | static void ext4_put_super(struct super_block *sb) |
ac27a0ec | 751 | { |
617ba13b MC |
752 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
753 | struct ext4_super_block *es = sbi->s_es; | |
ef2cabf7 | 754 | int i, err; |
ac27a0ec | 755 | |
857ac889 | 756 | ext4_unregister_li_request(sb); |
e0ccfd95 CH |
757 | dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED); |
758 | ||
4c0425ff MC |
759 | flush_workqueue(sbi->dio_unwritten_wq); |
760 | destroy_workqueue(sbi->dio_unwritten_wq); | |
761 | ||
a9e220f8 | 762 | lock_super(sb); |
8c85e125 | 763 | if (sb->s_dirt) |
ebc1ac16 | 764 | ext4_commit_super(sb, 1); |
8c85e125 | 765 | |
0390131b FM |
766 | if (sbi->s_journal) { |
767 | err = jbd2_journal_destroy(sbi->s_journal); | |
768 | sbi->s_journal = NULL; | |
769 | if (err < 0) | |
c67d859e | 770 | ext4_abort(sb, "Couldn't clean up the journal"); |
0390131b | 771 | } |
d4edac31 | 772 | |
a1c6c569 | 773 | del_timer(&sbi->s_err_report); |
d4edac31 JB |
774 | ext4_release_system_zone(sb); |
775 | ext4_mb_release(sb); | |
776 | ext4_ext_release(sb); | |
777 | ext4_xattr_put_super(sb); | |
778 | ||
ac27a0ec | 779 | if (!(sb->s_flags & MS_RDONLY)) { |
617ba13b | 780 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); |
ac27a0ec | 781 | es->s_state = cpu_to_le16(sbi->s_mount_state); |
e2d67052 | 782 | ext4_commit_super(sb, 1); |
ac27a0ec | 783 | } |
240799cd | 784 | if (sbi->s_proc) { |
9f6200bb | 785 | remove_proc_entry(sb->s_id, ext4_proc_root); |
240799cd | 786 | } |
3197ebdb | 787 | kobject_del(&sbi->s_kobj); |
ac27a0ec DK |
788 | |
789 | for (i = 0; i < sbi->s_gdb_count; i++) | |
790 | brelse(sbi->s_group_desc[i]); | |
791 | kfree(sbi->s_group_desc); | |
c5ca7c76 TT |
792 | if (is_vmalloc_addr(sbi->s_flex_groups)) |
793 | vfree(sbi->s_flex_groups); | |
794 | else | |
795 | kfree(sbi->s_flex_groups); | |
ac27a0ec DK |
796 | percpu_counter_destroy(&sbi->s_freeblocks_counter); |
797 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
798 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
6bc6e63f | 799 | percpu_counter_destroy(&sbi->s_dirtyblocks_counter); |
ac27a0ec DK |
800 | brelse(sbi->s_sbh); |
801 | #ifdef CONFIG_QUOTA | |
802 | for (i = 0; i < MAXQUOTAS; i++) | |
803 | kfree(sbi->s_qf_names[i]); | |
804 | #endif | |
805 | ||
806 | /* Debugging code just in case the in-memory inode orphan list | |
807 | * isn't empty. The on-disk one can be non-empty if we've | |
808 | * detected an error and taken the fs readonly, but the | |
809 | * in-memory list had better be clean by this point. */ | |
810 | if (!list_empty(&sbi->s_orphan)) | |
811 | dump_orphan_list(sb, sbi); | |
812 | J_ASSERT(list_empty(&sbi->s_orphan)); | |
813 | ||
f98393a6 | 814 | invalidate_bdev(sb->s_bdev); |
ac27a0ec DK |
815 | if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { |
816 | /* | |
817 | * Invalidate the journal device's buffers. We don't want them | |
818 | * floating about in memory - the physical journal device may | |
819 | * hotswapped, and it breaks the `ro-after' testing code. | |
820 | */ | |
821 | sync_blockdev(sbi->journal_bdev); | |
f98393a6 | 822 | invalidate_bdev(sbi->journal_bdev); |
617ba13b | 823 | ext4_blkdev_remove(sbi); |
ac27a0ec DK |
824 | } |
825 | sb->s_fs_info = NULL; | |
3197ebdb TT |
826 | /* |
827 | * Now that we are completely done shutting down the | |
828 | * superblock, we need to actually destroy the kobject. | |
829 | */ | |
3197ebdb TT |
830 | unlock_super(sb); |
831 | kobject_put(&sbi->s_kobj); | |
832 | wait_for_completion(&sbi->s_kobj_unregister); | |
705895b6 | 833 | kfree(sbi->s_blockgroup_lock); |
ac27a0ec | 834 | kfree(sbi); |
ac27a0ec DK |
835 | } |
836 | ||
e18b890b | 837 | static struct kmem_cache *ext4_inode_cachep; |
ac27a0ec DK |
838 | |
839 | /* | |
840 | * Called inside transaction, so use GFP_NOFS | |
841 | */ | |
617ba13b | 842 | static struct inode *ext4_alloc_inode(struct super_block *sb) |
ac27a0ec | 843 | { |
617ba13b | 844 | struct ext4_inode_info *ei; |
ac27a0ec | 845 | |
e6b4f8da | 846 | ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); |
ac27a0ec DK |
847 | if (!ei) |
848 | return NULL; | |
0b8e58a1 | 849 | |
ac27a0ec | 850 | ei->vfs_inode.i_version = 1; |
91246c00 | 851 | ei->vfs_inode.i_data.writeback_index = 0; |
a86c6181 | 852 | memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache)); |
c9de560d AT |
853 | INIT_LIST_HEAD(&ei->i_prealloc_list); |
854 | spin_lock_init(&ei->i_prealloc_lock); | |
d2a17637 MC |
855 | ei->i_reserved_data_blocks = 0; |
856 | ei->i_reserved_meta_blocks = 0; | |
857 | ei->i_allocated_meta_blocks = 0; | |
9d0be502 | 858 | ei->i_da_metadata_calc_len = 0; |
d2a17637 | 859 | spin_lock_init(&(ei->i_block_reservation_lock)); |
a9e7f447 DM |
860 | #ifdef CONFIG_QUOTA |
861 | ei->i_reserved_quota = 0; | |
862 | #endif | |
8aefcd55 | 863 | ei->jinode = NULL; |
c7064ef1 | 864 | INIT_LIST_HEAD(&ei->i_completed_io_list); |
744692dc | 865 | spin_lock_init(&ei->i_completed_io_lock); |
8d5d02e6 | 866 | ei->cur_aio_dio = NULL; |
b436b9be JK |
867 | ei->i_sync_tid = 0; |
868 | ei->i_datasync_tid = 0; | |
f7ad6d2e | 869 | atomic_set(&ei->i_ioend_count, 0); |
e9e3bcec | 870 | atomic_set(&ei->i_aiodio_unwritten, 0); |
0b8e58a1 | 871 | |
ac27a0ec DK |
872 | return &ei->vfs_inode; |
873 | } | |
874 | ||
7ff9c073 TT |
875 | static int ext4_drop_inode(struct inode *inode) |
876 | { | |
877 | int drop = generic_drop_inode(inode); | |
878 | ||
879 | trace_ext4_drop_inode(inode, drop); | |
880 | return drop; | |
881 | } | |
882 | ||
fa0d7e3d NP |
883 | static void ext4_i_callback(struct rcu_head *head) |
884 | { | |
885 | struct inode *inode = container_of(head, struct inode, i_rcu); | |
886 | INIT_LIST_HEAD(&inode->i_dentry); | |
887 | kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); | |
888 | } | |
889 | ||
617ba13b | 890 | static void ext4_destroy_inode(struct inode *inode) |
ac27a0ec | 891 | { |
f7ad6d2e | 892 | ext4_ioend_wait(inode); |
9f7dd93d | 893 | if (!list_empty(&(EXT4_I(inode)->i_orphan))) { |
b31e1552 ES |
894 | ext4_msg(inode->i_sb, KERN_ERR, |
895 | "Inode %lu (%p): orphan list check failed!", | |
896 | inode->i_ino, EXT4_I(inode)); | |
9f7dd93d VA |
897 | print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, |
898 | EXT4_I(inode), sizeof(struct ext4_inode_info), | |
899 | true); | |
900 | dump_stack(); | |
901 | } | |
fa0d7e3d | 902 | call_rcu(&inode->i_rcu, ext4_i_callback); |
ac27a0ec DK |
903 | } |
904 | ||
51cc5068 | 905 | static void init_once(void *foo) |
ac27a0ec | 906 | { |
617ba13b | 907 | struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; |
ac27a0ec | 908 | |
a35afb83 | 909 | INIT_LIST_HEAD(&ei->i_orphan); |
03010a33 | 910 | #ifdef CONFIG_EXT4_FS_XATTR |
a35afb83 | 911 | init_rwsem(&ei->xattr_sem); |
ac27a0ec | 912 | #endif |
0e855ac8 | 913 | init_rwsem(&ei->i_data_sem); |
a35afb83 | 914 | inode_init_once(&ei->vfs_inode); |
ac27a0ec DK |
915 | } |
916 | ||
917 | static int init_inodecache(void) | |
918 | { | |
617ba13b MC |
919 | ext4_inode_cachep = kmem_cache_create("ext4_inode_cache", |
920 | sizeof(struct ext4_inode_info), | |
ac27a0ec DK |
921 | 0, (SLAB_RECLAIM_ACCOUNT| |
922 | SLAB_MEM_SPREAD), | |
20c2df83 | 923 | init_once); |
617ba13b | 924 | if (ext4_inode_cachep == NULL) |
ac27a0ec DK |
925 | return -ENOMEM; |
926 | return 0; | |
927 | } | |
928 | ||
929 | static void destroy_inodecache(void) | |
930 | { | |
617ba13b | 931 | kmem_cache_destroy(ext4_inode_cachep); |
ac27a0ec DK |
932 | } |
933 | ||
0930fcc1 | 934 | void ext4_clear_inode(struct inode *inode) |
ac27a0ec | 935 | { |
0930fcc1 AV |
936 | invalidate_inode_buffers(inode); |
937 | end_writeback(inode); | |
9f754758 | 938 | dquot_drop(inode); |
c2ea3fde | 939 | ext4_discard_preallocations(inode); |
8aefcd55 TT |
940 | if (EXT4_I(inode)->jinode) { |
941 | jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), | |
942 | EXT4_I(inode)->jinode); | |
943 | jbd2_free_inode(EXT4_I(inode)->jinode); | |
944 | EXT4_I(inode)->jinode = NULL; | |
945 | } | |
ac27a0ec DK |
946 | } |
947 | ||
2b2d6d01 TT |
948 | static inline void ext4_show_quota_options(struct seq_file *seq, |
949 | struct super_block *sb) | |
ac27a0ec DK |
950 | { |
951 | #if defined(CONFIG_QUOTA) | |
617ba13b | 952 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 953 | |
5a20bdfc JK |
954 | if (sbi->s_jquota_fmt) { |
955 | char *fmtname = ""; | |
956 | ||
957 | switch (sbi->s_jquota_fmt) { | |
958 | case QFMT_VFS_OLD: | |
959 | fmtname = "vfsold"; | |
960 | break; | |
961 | case QFMT_VFS_V0: | |
962 | fmtname = "vfsv0"; | |
963 | break; | |
964 | case QFMT_VFS_V1: | |
965 | fmtname = "vfsv1"; | |
966 | break; | |
967 | } | |
968 | seq_printf(seq, ",jqfmt=%s", fmtname); | |
969 | } | |
ac27a0ec DK |
970 | |
971 | if (sbi->s_qf_names[USRQUOTA]) | |
972 | seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]); | |
973 | ||
974 | if (sbi->s_qf_names[GRPQUOTA]) | |
975 | seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]); | |
976 | ||
482a7425 | 977 | if (test_opt(sb, USRQUOTA)) |
ac27a0ec DK |
978 | seq_puts(seq, ",usrquota"); |
979 | ||
482a7425 | 980 | if (test_opt(sb, GRPQUOTA)) |
ac27a0ec DK |
981 | seq_puts(seq, ",grpquota"); |
982 | #endif | |
983 | } | |
984 | ||
d9c9bef1 MS |
985 | /* |
986 | * Show an option if | |
987 | * - it's set to a non-default value OR | |
988 | * - if the per-sb default is different from the global default | |
989 | */ | |
617ba13b | 990 | static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs) |
ac27a0ec | 991 | { |
aa22df2c AK |
992 | int def_errors; |
993 | unsigned long def_mount_opts; | |
ac27a0ec | 994 | struct super_block *sb = vfs->mnt_sb; |
d9c9bef1 MS |
995 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
996 | struct ext4_super_block *es = sbi->s_es; | |
d9c9bef1 MS |
997 | |
998 | def_mount_opts = le32_to_cpu(es->s_default_mount_opts); | |
aa22df2c | 999 | def_errors = le16_to_cpu(es->s_errors); |
d9c9bef1 MS |
1000 | |
1001 | if (sbi->s_sb_block != 1) | |
1002 | seq_printf(seq, ",sb=%llu", sbi->s_sb_block); | |
1003 | if (test_opt(sb, MINIX_DF)) | |
1004 | seq_puts(seq, ",minixdf"); | |
aa22df2c | 1005 | if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS)) |
d9c9bef1 MS |
1006 | seq_puts(seq, ",grpid"); |
1007 | if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS)) | |
1008 | seq_puts(seq, ",nogrpid"); | |
1009 | if (sbi->s_resuid != EXT4_DEF_RESUID || | |
1010 | le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) { | |
1011 | seq_printf(seq, ",resuid=%u", sbi->s_resuid); | |
1012 | } | |
1013 | if (sbi->s_resgid != EXT4_DEF_RESGID || | |
1014 | le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) { | |
1015 | seq_printf(seq, ",resgid=%u", sbi->s_resgid); | |
1016 | } | |
bb4f397a | 1017 | if (test_opt(sb, ERRORS_RO)) { |
d9c9bef1 | 1018 | if (def_errors == EXT4_ERRORS_PANIC || |
bb4f397a AK |
1019 | def_errors == EXT4_ERRORS_CONTINUE) { |
1020 | seq_puts(seq, ",errors=remount-ro"); | |
d9c9bef1 MS |
1021 | } |
1022 | } | |
aa22df2c | 1023 | if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) |
bb4f397a | 1024 | seq_puts(seq, ",errors=continue"); |
aa22df2c | 1025 | if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) |
d9c9bef1 | 1026 | seq_puts(seq, ",errors=panic"); |
aa22df2c | 1027 | if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16)) |
d9c9bef1 | 1028 | seq_puts(seq, ",nouid32"); |
aa22df2c | 1029 | if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG)) |
d9c9bef1 MS |
1030 | seq_puts(seq, ",debug"); |
1031 | if (test_opt(sb, OLDALLOC)) | |
1032 | seq_puts(seq, ",oldalloc"); | |
03010a33 | 1033 | #ifdef CONFIG_EXT4_FS_XATTR |
ea663336 | 1034 | if (test_opt(sb, XATTR_USER)) |
d9c9bef1 | 1035 | seq_puts(seq, ",user_xattr"); |
ea663336 | 1036 | if (!test_opt(sb, XATTR_USER)) |
d9c9bef1 | 1037 | seq_puts(seq, ",nouser_xattr"); |
d9c9bef1 | 1038 | #endif |
03010a33 | 1039 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
aa22df2c | 1040 | if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL)) |
d9c9bef1 MS |
1041 | seq_puts(seq, ",acl"); |
1042 | if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL)) | |
1043 | seq_puts(seq, ",noacl"); | |
1044 | #endif | |
30773840 | 1045 | if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { |
d9c9bef1 MS |
1046 | seq_printf(seq, ",commit=%u", |
1047 | (unsigned) (sbi->s_commit_interval / HZ)); | |
1048 | } | |
30773840 TT |
1049 | if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) { |
1050 | seq_printf(seq, ",min_batch_time=%u", | |
1051 | (unsigned) sbi->s_min_batch_time); | |
1052 | } | |
1053 | if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) { | |
1054 | seq_printf(seq, ",max_batch_time=%u", | |
1055 | (unsigned) sbi->s_min_batch_time); | |
1056 | } | |
1057 | ||
571640ca ES |
1058 | /* |
1059 | * We're changing the default of barrier mount option, so | |
1060 | * let's always display its mount state so it's clear what its | |
1061 | * status is. | |
1062 | */ | |
1063 | seq_puts(seq, ",barrier="); | |
1064 | seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0"); | |
cd0b6a39 TT |
1065 | if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) |
1066 | seq_puts(seq, ",journal_async_commit"); | |
39a4bade JK |
1067 | else if (test_opt(sb, JOURNAL_CHECKSUM)) |
1068 | seq_puts(seq, ",journal_checksum"); | |
25ec56b5 JNC |
1069 | if (test_opt(sb, I_VERSION)) |
1070 | seq_puts(seq, ",i_version"); | |
8b67f04a TT |
1071 | if (!test_opt(sb, DELALLOC) && |
1072 | !(def_mount_opts & EXT4_DEFM_NODELALLOC)) | |
dd919b98 AK |
1073 | seq_puts(seq, ",nodelalloc"); |
1074 | ||
6fd7a467 TT |
1075 | if (!test_opt(sb, MBLK_IO_SUBMIT)) |
1076 | seq_puts(seq, ",nomblk_io_submit"); | |
cb45bbe4 MS |
1077 | if (sbi->s_stripe) |
1078 | seq_printf(seq, ",stripe=%lu", sbi->s_stripe); | |
aa22df2c AK |
1079 | /* |
1080 | * journal mode get enabled in different ways | |
1081 | * So just print the value even if we didn't specify it | |
1082 | */ | |
617ba13b | 1083 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) |
ac27a0ec | 1084 | seq_puts(seq, ",data=journal"); |
617ba13b | 1085 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) |
ac27a0ec | 1086 | seq_puts(seq, ",data=ordered"); |
617ba13b | 1087 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) |
ac27a0ec DK |
1088 | seq_puts(seq, ",data=writeback"); |
1089 | ||
240799cd TT |
1090 | if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) |
1091 | seq_printf(seq, ",inode_readahead_blks=%u", | |
1092 | sbi->s_inode_readahead_blks); | |
1093 | ||
5bf5683a HK |
1094 | if (test_opt(sb, DATA_ERR_ABORT)) |
1095 | seq_puts(seq, ",data_err=abort"); | |
1096 | ||
afd4672d | 1097 | if (test_opt(sb, NO_AUTO_DA_ALLOC)) |
06705bff | 1098 | seq_puts(seq, ",noauto_da_alloc"); |
afd4672d | 1099 | |
8b67f04a | 1100 | if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD)) |
5328e635 ES |
1101 | seq_puts(seq, ",discard"); |
1102 | ||
e3bb52ae ES |
1103 | if (test_opt(sb, NOLOAD)) |
1104 | seq_puts(seq, ",norecovery"); | |
1105 | ||
744692dc JZ |
1106 | if (test_opt(sb, DIOREAD_NOLOCK)) |
1107 | seq_puts(seq, ",dioread_nolock"); | |
1108 | ||
8b67f04a TT |
1109 | if (test_opt(sb, BLOCK_VALIDITY) && |
1110 | !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)) | |
1111 | seq_puts(seq, ",block_validity"); | |
1112 | ||
bfff6873 LC |
1113 | if (!test_opt(sb, INIT_INODE_TABLE)) |
1114 | seq_puts(seq, ",noinit_inode_table"); | |
51ce6511 | 1115 | else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT) |
bfff6873 LC |
1116 | seq_printf(seq, ",init_inode_table=%u", |
1117 | (unsigned) sbi->s_li_wait_mult); | |
1118 | ||
617ba13b | 1119 | ext4_show_quota_options(seq, sb); |
0b8e58a1 | 1120 | |
ac27a0ec DK |
1121 | return 0; |
1122 | } | |
1123 | ||
1b961ac0 | 1124 | static struct inode *ext4_nfs_get_inode(struct super_block *sb, |
0b8e58a1 | 1125 | u64 ino, u32 generation) |
ac27a0ec | 1126 | { |
ac27a0ec | 1127 | struct inode *inode; |
ac27a0ec | 1128 | |
617ba13b | 1129 | if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO) |
ac27a0ec | 1130 | return ERR_PTR(-ESTALE); |
617ba13b | 1131 | if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count)) |
ac27a0ec DK |
1132 | return ERR_PTR(-ESTALE); |
1133 | ||
1134 | /* iget isn't really right if the inode is currently unallocated!! | |
1135 | * | |
617ba13b | 1136 | * ext4_read_inode will return a bad_inode if the inode had been |
ac27a0ec DK |
1137 | * deleted, so we should be safe. |
1138 | * | |
1139 | * Currently we don't know the generation for parent directory, so | |
1140 | * a generation of 0 means "accept any" | |
1141 | */ | |
1d1fe1ee DH |
1142 | inode = ext4_iget(sb, ino); |
1143 | if (IS_ERR(inode)) | |
1144 | return ERR_CAST(inode); | |
1145 | if (generation && inode->i_generation != generation) { | |
ac27a0ec DK |
1146 | iput(inode); |
1147 | return ERR_PTR(-ESTALE); | |
1148 | } | |
1b961ac0 CH |
1149 | |
1150 | return inode; | |
1151 | } | |
1152 | ||
1153 | static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, | |
0b8e58a1 | 1154 | int fh_len, int fh_type) |
1b961ac0 CH |
1155 | { |
1156 | return generic_fh_to_dentry(sb, fid, fh_len, fh_type, | |
1157 | ext4_nfs_get_inode); | |
1158 | } | |
1159 | ||
1160 | static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, | |
0b8e58a1 | 1161 | int fh_len, int fh_type) |
1b961ac0 CH |
1162 | { |
1163 | return generic_fh_to_parent(sb, fid, fh_len, fh_type, | |
1164 | ext4_nfs_get_inode); | |
ac27a0ec DK |
1165 | } |
1166 | ||
c39a7f84 TO |
1167 | /* |
1168 | * Try to release metadata pages (indirect blocks, directories) which are | |
1169 | * mapped via the block device. Since these pages could have journal heads | |
1170 | * which would prevent try_to_free_buffers() from freeing them, we must use | |
1171 | * jbd2 layer's try_to_free_buffers() function to release them. | |
1172 | */ | |
0b8e58a1 AD |
1173 | static int bdev_try_to_free_page(struct super_block *sb, struct page *page, |
1174 | gfp_t wait) | |
c39a7f84 TO |
1175 | { |
1176 | journal_t *journal = EXT4_SB(sb)->s_journal; | |
1177 | ||
1178 | WARN_ON(PageChecked(page)); | |
1179 | if (!page_has_buffers(page)) | |
1180 | return 0; | |
1181 | if (journal) | |
1182 | return jbd2_journal_try_to_free_buffers(journal, page, | |
1183 | wait & ~__GFP_WAIT); | |
1184 | return try_to_free_buffers(page); | |
1185 | } | |
1186 | ||
ac27a0ec | 1187 | #ifdef CONFIG_QUOTA |
af5bc92d | 1188 | #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group") |
2b2d6d01 | 1189 | #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA)) |
ac27a0ec | 1190 | |
617ba13b MC |
1191 | static int ext4_write_dquot(struct dquot *dquot); |
1192 | static int ext4_acquire_dquot(struct dquot *dquot); | |
1193 | static int ext4_release_dquot(struct dquot *dquot); | |
1194 | static int ext4_mark_dquot_dirty(struct dquot *dquot); | |
1195 | static int ext4_write_info(struct super_block *sb, int type); | |
6f28e087 | 1196 | static int ext4_quota_on(struct super_block *sb, int type, int format_id, |
f00c9e44 | 1197 | struct path *path); |
ca0e05e4 | 1198 | static int ext4_quota_off(struct super_block *sb, int type); |
617ba13b MC |
1199 | static int ext4_quota_on_mount(struct super_block *sb, int type); |
1200 | static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, | |
ac27a0ec | 1201 | size_t len, loff_t off); |
617ba13b | 1202 | static ssize_t ext4_quota_write(struct super_block *sb, int type, |
ac27a0ec DK |
1203 | const char *data, size_t len, loff_t off); |
1204 | ||
61e225dc | 1205 | static const struct dquot_operations ext4_quota_operations = { |
60e58e0f | 1206 | .get_reserved_space = ext4_get_reserved_space, |
617ba13b MC |
1207 | .write_dquot = ext4_write_dquot, |
1208 | .acquire_dquot = ext4_acquire_dquot, | |
1209 | .release_dquot = ext4_release_dquot, | |
1210 | .mark_dirty = ext4_mark_dquot_dirty, | |
a5b5ee32 JK |
1211 | .write_info = ext4_write_info, |
1212 | .alloc_dquot = dquot_alloc, | |
1213 | .destroy_dquot = dquot_destroy, | |
ac27a0ec DK |
1214 | }; |
1215 | ||
0d54b217 | 1216 | static const struct quotactl_ops ext4_qctl_operations = { |
617ba13b | 1217 | .quota_on = ext4_quota_on, |
ca0e05e4 | 1218 | .quota_off = ext4_quota_off, |
287a8095 CH |
1219 | .quota_sync = dquot_quota_sync, |
1220 | .get_info = dquot_get_dqinfo, | |
1221 | .set_info = dquot_set_dqinfo, | |
1222 | .get_dqblk = dquot_get_dqblk, | |
1223 | .set_dqblk = dquot_set_dqblk | |
ac27a0ec DK |
1224 | }; |
1225 | #endif | |
1226 | ||
ee9b6d61 | 1227 | static const struct super_operations ext4_sops = { |
617ba13b MC |
1228 | .alloc_inode = ext4_alloc_inode, |
1229 | .destroy_inode = ext4_destroy_inode, | |
617ba13b MC |
1230 | .write_inode = ext4_write_inode, |
1231 | .dirty_inode = ext4_dirty_inode, | |
7ff9c073 | 1232 | .drop_inode = ext4_drop_inode, |
0930fcc1 | 1233 | .evict_inode = ext4_evict_inode, |
617ba13b | 1234 | .put_super = ext4_put_super, |
617ba13b | 1235 | .sync_fs = ext4_sync_fs, |
c4be0c1d TS |
1236 | .freeze_fs = ext4_freeze, |
1237 | .unfreeze_fs = ext4_unfreeze, | |
617ba13b MC |
1238 | .statfs = ext4_statfs, |
1239 | .remount_fs = ext4_remount, | |
617ba13b | 1240 | .show_options = ext4_show_options, |
ac27a0ec | 1241 | #ifdef CONFIG_QUOTA |
617ba13b MC |
1242 | .quota_read = ext4_quota_read, |
1243 | .quota_write = ext4_quota_write, | |
ac27a0ec | 1244 | #endif |
c39a7f84 | 1245 | .bdev_try_to_free_page = bdev_try_to_free_page, |
ac27a0ec DK |
1246 | }; |
1247 | ||
9ca92389 TT |
1248 | static const struct super_operations ext4_nojournal_sops = { |
1249 | .alloc_inode = ext4_alloc_inode, | |
1250 | .destroy_inode = ext4_destroy_inode, | |
1251 | .write_inode = ext4_write_inode, | |
1252 | .dirty_inode = ext4_dirty_inode, | |
7ff9c073 | 1253 | .drop_inode = ext4_drop_inode, |
0930fcc1 | 1254 | .evict_inode = ext4_evict_inode, |
9ca92389 TT |
1255 | .write_super = ext4_write_super, |
1256 | .put_super = ext4_put_super, | |
1257 | .statfs = ext4_statfs, | |
1258 | .remount_fs = ext4_remount, | |
9ca92389 TT |
1259 | .show_options = ext4_show_options, |
1260 | #ifdef CONFIG_QUOTA | |
1261 | .quota_read = ext4_quota_read, | |
1262 | .quota_write = ext4_quota_write, | |
1263 | #endif | |
1264 | .bdev_try_to_free_page = bdev_try_to_free_page, | |
1265 | }; | |
1266 | ||
39655164 | 1267 | static const struct export_operations ext4_export_ops = { |
1b961ac0 CH |
1268 | .fh_to_dentry = ext4_fh_to_dentry, |
1269 | .fh_to_parent = ext4_fh_to_parent, | |
617ba13b | 1270 | .get_parent = ext4_get_parent, |
ac27a0ec DK |
1271 | }; |
1272 | ||
1273 | enum { | |
1274 | Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, | |
1275 | Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, | |
01436ef2 | 1276 | Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov, |
ac27a0ec | 1277 | Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, |
06705bff | 1278 | Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh, |
30773840 | 1279 | Opt_commit, Opt_min_batch_time, Opt_max_batch_time, |
c3191067 | 1280 | Opt_journal_update, Opt_journal_dev, |
818d276c | 1281 | Opt_journal_checksum, Opt_journal_async_commit, |
ac27a0ec | 1282 | Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, |
296c355c | 1283 | Opt_data_err_abort, Opt_data_err_ignore, |
ac27a0ec | 1284 | Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, |
5a20bdfc JK |
1285 | Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, |
1286 | Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, | |
1287 | Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version, | |
1449032b TT |
1288 | Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit, |
1289 | Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, | |
5328e635 | 1290 | Opt_inode_readahead_blks, Opt_journal_ioprio, |
744692dc | 1291 | Opt_dioread_nolock, Opt_dioread_lock, |
5328e635 | 1292 | Opt_discard, Opt_nodiscard, |
bfff6873 | 1293 | Opt_init_inode_table, Opt_noinit_inode_table, |
ac27a0ec DK |
1294 | }; |
1295 | ||
a447c093 | 1296 | static const match_table_t tokens = { |
ac27a0ec DK |
1297 | {Opt_bsd_df, "bsddf"}, |
1298 | {Opt_minix_df, "minixdf"}, | |
1299 | {Opt_grpid, "grpid"}, | |
1300 | {Opt_grpid, "bsdgroups"}, | |
1301 | {Opt_nogrpid, "nogrpid"}, | |
1302 | {Opt_nogrpid, "sysvgroups"}, | |
1303 | {Opt_resgid, "resgid=%u"}, | |
1304 | {Opt_resuid, "resuid=%u"}, | |
1305 | {Opt_sb, "sb=%u"}, | |
1306 | {Opt_err_cont, "errors=continue"}, | |
1307 | {Opt_err_panic, "errors=panic"}, | |
1308 | {Opt_err_ro, "errors=remount-ro"}, | |
1309 | {Opt_nouid32, "nouid32"}, | |
ac27a0ec DK |
1310 | {Opt_debug, "debug"}, |
1311 | {Opt_oldalloc, "oldalloc"}, | |
1312 | {Opt_orlov, "orlov"}, | |
1313 | {Opt_user_xattr, "user_xattr"}, | |
1314 | {Opt_nouser_xattr, "nouser_xattr"}, | |
1315 | {Opt_acl, "acl"}, | |
1316 | {Opt_noacl, "noacl"}, | |
ac27a0ec | 1317 | {Opt_noload, "noload"}, |
e3bb52ae | 1318 | {Opt_noload, "norecovery"}, |
ac27a0ec DK |
1319 | {Opt_nobh, "nobh"}, |
1320 | {Opt_bh, "bh"}, | |
1321 | {Opt_commit, "commit=%u"}, | |
30773840 TT |
1322 | {Opt_min_batch_time, "min_batch_time=%u"}, |
1323 | {Opt_max_batch_time, "max_batch_time=%u"}, | |
ac27a0ec | 1324 | {Opt_journal_update, "journal=update"}, |
ac27a0ec | 1325 | {Opt_journal_dev, "journal_dev=%u"}, |
818d276c GS |
1326 | {Opt_journal_checksum, "journal_checksum"}, |
1327 | {Opt_journal_async_commit, "journal_async_commit"}, | |
ac27a0ec DK |
1328 | {Opt_abort, "abort"}, |
1329 | {Opt_data_journal, "data=journal"}, | |
1330 | {Opt_data_ordered, "data=ordered"}, | |
1331 | {Opt_data_writeback, "data=writeback"}, | |
5bf5683a HK |
1332 | {Opt_data_err_abort, "data_err=abort"}, |
1333 | {Opt_data_err_ignore, "data_err=ignore"}, | |
ac27a0ec DK |
1334 | {Opt_offusrjquota, "usrjquota="}, |
1335 | {Opt_usrjquota, "usrjquota=%s"}, | |
1336 | {Opt_offgrpjquota, "grpjquota="}, | |
1337 | {Opt_grpjquota, "grpjquota=%s"}, | |
1338 | {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, | |
1339 | {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, | |
5a20bdfc | 1340 | {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, |
ac27a0ec DK |
1341 | {Opt_grpquota, "grpquota"}, |
1342 | {Opt_noquota, "noquota"}, | |
1343 | {Opt_quota, "quota"}, | |
1344 | {Opt_usrquota, "usrquota"}, | |
1345 | {Opt_barrier, "barrier=%u"}, | |
06705bff TT |
1346 | {Opt_barrier, "barrier"}, |
1347 | {Opt_nobarrier, "nobarrier"}, | |
25ec56b5 | 1348 | {Opt_i_version, "i_version"}, |
c9de560d | 1349 | {Opt_stripe, "stripe=%u"}, |
ac27a0ec | 1350 | {Opt_resize, "resize"}, |
64769240 | 1351 | {Opt_delalloc, "delalloc"}, |
dd919b98 | 1352 | {Opt_nodelalloc, "nodelalloc"}, |
1449032b TT |
1353 | {Opt_mblk_io_submit, "mblk_io_submit"}, |
1354 | {Opt_nomblk_io_submit, "nomblk_io_submit"}, | |
6fd058f7 TT |
1355 | {Opt_block_validity, "block_validity"}, |
1356 | {Opt_noblock_validity, "noblock_validity"}, | |
240799cd | 1357 | {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, |
b3881f74 | 1358 | {Opt_journal_ioprio, "journal_ioprio=%u"}, |
afd4672d | 1359 | {Opt_auto_da_alloc, "auto_da_alloc=%u"}, |
06705bff TT |
1360 | {Opt_auto_da_alloc, "auto_da_alloc"}, |
1361 | {Opt_noauto_da_alloc, "noauto_da_alloc"}, | |
744692dc JZ |
1362 | {Opt_dioread_nolock, "dioread_nolock"}, |
1363 | {Opt_dioread_lock, "dioread_lock"}, | |
5328e635 ES |
1364 | {Opt_discard, "discard"}, |
1365 | {Opt_nodiscard, "nodiscard"}, | |
bfff6873 LC |
1366 | {Opt_init_inode_table, "init_itable=%u"}, |
1367 | {Opt_init_inode_table, "init_itable"}, | |
1368 | {Opt_noinit_inode_table, "noinit_itable"}, | |
f3f12faa | 1369 | {Opt_err, NULL}, |
ac27a0ec DK |
1370 | }; |
1371 | ||
617ba13b | 1372 | static ext4_fsblk_t get_sb_block(void **data) |
ac27a0ec | 1373 | { |
617ba13b | 1374 | ext4_fsblk_t sb_block; |
ac27a0ec DK |
1375 | char *options = (char *) *data; |
1376 | ||
1377 | if (!options || strncmp(options, "sb=", 3) != 0) | |
1378 | return 1; /* Default location */ | |
0b8e58a1 | 1379 | |
ac27a0ec | 1380 | options += 3; |
0b8e58a1 | 1381 | /* TODO: use simple_strtoll with >32bit ext4 */ |
ac27a0ec DK |
1382 | sb_block = simple_strtoul(options, &options, 0); |
1383 | if (*options && *options != ',') { | |
4776004f | 1384 | printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", |
ac27a0ec DK |
1385 | (char *) *data); |
1386 | return 1; | |
1387 | } | |
1388 | if (*options == ',') | |
1389 | options++; | |
1390 | *data = (void *) options; | |
0b8e58a1 | 1391 | |
ac27a0ec DK |
1392 | return sb_block; |
1393 | } | |
1394 | ||
b3881f74 | 1395 | #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) |
437ca0fd DM |
1396 | static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n" |
1397 | "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; | |
b3881f74 | 1398 | |
56c50f11 DM |
1399 | #ifdef CONFIG_QUOTA |
1400 | static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) | |
1401 | { | |
1402 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1403 | char *qname; | |
1404 | ||
1405 | if (sb_any_quota_loaded(sb) && | |
1406 | !sbi->s_qf_names[qtype]) { | |
1407 | ext4_msg(sb, KERN_ERR, | |
1408 | "Cannot change journaled " | |
1409 | "quota options when quota turned on"); | |
1410 | return 0; | |
1411 | } | |
1412 | qname = match_strdup(args); | |
1413 | if (!qname) { | |
1414 | ext4_msg(sb, KERN_ERR, | |
1415 | "Not enough memory for storing quotafile name"); | |
1416 | return 0; | |
1417 | } | |
1418 | if (sbi->s_qf_names[qtype] && | |
1419 | strcmp(sbi->s_qf_names[qtype], qname)) { | |
1420 | ext4_msg(sb, KERN_ERR, | |
1421 | "%s quota file already specified", QTYPE2NAME(qtype)); | |
1422 | kfree(qname); | |
1423 | return 0; | |
1424 | } | |
1425 | sbi->s_qf_names[qtype] = qname; | |
1426 | if (strchr(sbi->s_qf_names[qtype], '/')) { | |
1427 | ext4_msg(sb, KERN_ERR, | |
1428 | "quotafile must be on filesystem root"); | |
1429 | kfree(sbi->s_qf_names[qtype]); | |
1430 | sbi->s_qf_names[qtype] = NULL; | |
1431 | return 0; | |
1432 | } | |
fd8c37ec | 1433 | set_opt(sb, QUOTA); |
56c50f11 DM |
1434 | return 1; |
1435 | } | |
1436 | ||
1437 | static int clear_qf_name(struct super_block *sb, int qtype) | |
1438 | { | |
1439 | ||
1440 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1441 | ||
1442 | if (sb_any_quota_loaded(sb) && | |
1443 | sbi->s_qf_names[qtype]) { | |
1444 | ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" | |
1445 | " when quota turned on"); | |
1446 | return 0; | |
1447 | } | |
1448 | /* | |
1449 | * The space will be released later when all options are confirmed | |
1450 | * to be correct | |
1451 | */ | |
1452 | sbi->s_qf_names[qtype] = NULL; | |
1453 | return 1; | |
1454 | } | |
1455 | #endif | |
1456 | ||
2b2d6d01 | 1457 | static int parse_options(char *options, struct super_block *sb, |
c3191067 | 1458 | unsigned long *journal_devnum, |
b3881f74 | 1459 | unsigned int *journal_ioprio, |
2b2d6d01 | 1460 | ext4_fsblk_t *n_blocks_count, int is_remount) |
ac27a0ec | 1461 | { |
617ba13b | 1462 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
2b2d6d01 | 1463 | char *p; |
ac27a0ec DK |
1464 | substring_t args[MAX_OPT_ARGS]; |
1465 | int data_opt = 0; | |
1466 | int option; | |
1467 | #ifdef CONFIG_QUOTA | |
56c50f11 | 1468 | int qfmt; |
ac27a0ec DK |
1469 | #endif |
1470 | ||
1471 | if (!options) | |
1472 | return 1; | |
1473 | ||
2b2d6d01 | 1474 | while ((p = strsep(&options, ",")) != NULL) { |
ac27a0ec DK |
1475 | int token; |
1476 | if (!*p) | |
1477 | continue; | |
1478 | ||
15121c18 ES |
1479 | /* |
1480 | * Initialize args struct so we know whether arg was | |
1481 | * found; some options take optional arguments. | |
1482 | */ | |
7dc57615 | 1483 | args[0].to = args[0].from = NULL; |
ac27a0ec DK |
1484 | token = match_token(p, tokens, args); |
1485 | switch (token) { | |
1486 | case Opt_bsd_df: | |
437ca0fd | 1487 | ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); |
fd8c37ec | 1488 | clear_opt(sb, MINIX_DF); |
ac27a0ec DK |
1489 | break; |
1490 | case Opt_minix_df: | |
437ca0fd | 1491 | ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); |
fd8c37ec | 1492 | set_opt(sb, MINIX_DF); |
437ca0fd | 1493 | |
ac27a0ec DK |
1494 | break; |
1495 | case Opt_grpid: | |
437ca0fd | 1496 | ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); |
fd8c37ec | 1497 | set_opt(sb, GRPID); |
437ca0fd | 1498 | |
ac27a0ec DK |
1499 | break; |
1500 | case Opt_nogrpid: | |
437ca0fd | 1501 | ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); |
fd8c37ec | 1502 | clear_opt(sb, GRPID); |
437ca0fd | 1503 | |
ac27a0ec DK |
1504 | break; |
1505 | case Opt_resuid: | |
1506 | if (match_int(&args[0], &option)) | |
1507 | return 0; | |
1508 | sbi->s_resuid = option; | |
1509 | break; | |
1510 | case Opt_resgid: | |
1511 | if (match_int(&args[0], &option)) | |
1512 | return 0; | |
1513 | sbi->s_resgid = option; | |
1514 | break; | |
1515 | case Opt_sb: | |
1516 | /* handled by get_sb_block() instead of here */ | |
1517 | /* *sb_block = match_int(&args[0]); */ | |
1518 | break; | |
1519 | case Opt_err_panic: | |
fd8c37ec TT |
1520 | clear_opt(sb, ERRORS_CONT); |
1521 | clear_opt(sb, ERRORS_RO); | |
1522 | set_opt(sb, ERRORS_PANIC); | |
ac27a0ec DK |
1523 | break; |
1524 | case Opt_err_ro: | |
fd8c37ec TT |
1525 | clear_opt(sb, ERRORS_CONT); |
1526 | clear_opt(sb, ERRORS_PANIC); | |
1527 | set_opt(sb, ERRORS_RO); | |
ac27a0ec DK |
1528 | break; |
1529 | case Opt_err_cont: | |
fd8c37ec TT |
1530 | clear_opt(sb, ERRORS_RO); |
1531 | clear_opt(sb, ERRORS_PANIC); | |
1532 | set_opt(sb, ERRORS_CONT); | |
ac27a0ec DK |
1533 | break; |
1534 | case Opt_nouid32: | |
fd8c37ec | 1535 | set_opt(sb, NO_UID32); |
ac27a0ec | 1536 | break; |
ac27a0ec | 1537 | case Opt_debug: |
fd8c37ec | 1538 | set_opt(sb, DEBUG); |
ac27a0ec DK |
1539 | break; |
1540 | case Opt_oldalloc: | |
fd8c37ec | 1541 | set_opt(sb, OLDALLOC); |
ac27a0ec DK |
1542 | break; |
1543 | case Opt_orlov: | |
fd8c37ec | 1544 | clear_opt(sb, OLDALLOC); |
ac27a0ec | 1545 | break; |
03010a33 | 1546 | #ifdef CONFIG_EXT4_FS_XATTR |
ac27a0ec | 1547 | case Opt_user_xattr: |
fd8c37ec | 1548 | set_opt(sb, XATTR_USER); |
ac27a0ec DK |
1549 | break; |
1550 | case Opt_nouser_xattr: | |
fd8c37ec | 1551 | clear_opt(sb, XATTR_USER); |
ac27a0ec DK |
1552 | break; |
1553 | #else | |
1554 | case Opt_user_xattr: | |
1555 | case Opt_nouser_xattr: | |
b31e1552 | 1556 | ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported"); |
ac27a0ec DK |
1557 | break; |
1558 | #endif | |
03010a33 | 1559 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
ac27a0ec | 1560 | case Opt_acl: |
fd8c37ec | 1561 | set_opt(sb, POSIX_ACL); |
ac27a0ec DK |
1562 | break; |
1563 | case Opt_noacl: | |
fd8c37ec | 1564 | clear_opt(sb, POSIX_ACL); |
ac27a0ec DK |
1565 | break; |
1566 | #else | |
1567 | case Opt_acl: | |
1568 | case Opt_noacl: | |
b31e1552 | 1569 | ext4_msg(sb, KERN_ERR, "(no)acl options not supported"); |
ac27a0ec DK |
1570 | break; |
1571 | #endif | |
ac27a0ec DK |
1572 | case Opt_journal_update: |
1573 | /* @@@ FIXME */ | |
1574 | /* Eventually we will want to be able to create | |
1575 | a journal file here. For now, only allow the | |
1576 | user to specify an existing inode to be the | |
1577 | journal file. */ | |
1578 | if (is_remount) { | |
b31e1552 ES |
1579 | ext4_msg(sb, KERN_ERR, |
1580 | "Cannot specify journal on remount"); | |
ac27a0ec DK |
1581 | return 0; |
1582 | } | |
fd8c37ec | 1583 | set_opt(sb, UPDATE_JOURNAL); |
ac27a0ec | 1584 | break; |
ac27a0ec DK |
1585 | case Opt_journal_dev: |
1586 | if (is_remount) { | |
b31e1552 ES |
1587 | ext4_msg(sb, KERN_ERR, |
1588 | "Cannot specify journal on remount"); | |
ac27a0ec DK |
1589 | return 0; |
1590 | } | |
1591 | if (match_int(&args[0], &option)) | |
1592 | return 0; | |
1593 | *journal_devnum = option; | |
1594 | break; | |
818d276c | 1595 | case Opt_journal_checksum: |
fd8c37ec | 1596 | set_opt(sb, JOURNAL_CHECKSUM); |
d4da6c9c | 1597 | break; |
818d276c | 1598 | case Opt_journal_async_commit: |
fd8c37ec TT |
1599 | set_opt(sb, JOURNAL_ASYNC_COMMIT); |
1600 | set_opt(sb, JOURNAL_CHECKSUM); | |
818d276c | 1601 | break; |
ac27a0ec | 1602 | case Opt_noload: |
fd8c37ec | 1603 | set_opt(sb, NOLOAD); |
ac27a0ec DK |
1604 | break; |
1605 | case Opt_commit: | |
1606 | if (match_int(&args[0], &option)) | |
1607 | return 0; | |
1608 | if (option < 0) | |
1609 | return 0; | |
1610 | if (option == 0) | |
cd02ff0b | 1611 | option = JBD2_DEFAULT_MAX_COMMIT_AGE; |
ac27a0ec DK |
1612 | sbi->s_commit_interval = HZ * option; |
1613 | break; | |
30773840 TT |
1614 | case Opt_max_batch_time: |
1615 | if (match_int(&args[0], &option)) | |
1616 | return 0; | |
1617 | if (option < 0) | |
1618 | return 0; | |
1619 | if (option == 0) | |
1620 | option = EXT4_DEF_MAX_BATCH_TIME; | |
1621 | sbi->s_max_batch_time = option; | |
1622 | break; | |
1623 | case Opt_min_batch_time: | |
1624 | if (match_int(&args[0], &option)) | |
1625 | return 0; | |
1626 | if (option < 0) | |
1627 | return 0; | |
1628 | sbi->s_min_batch_time = option; | |
1629 | break; | |
ac27a0ec | 1630 | case Opt_data_journal: |
617ba13b | 1631 | data_opt = EXT4_MOUNT_JOURNAL_DATA; |
ac27a0ec DK |
1632 | goto datacheck; |
1633 | case Opt_data_ordered: | |
617ba13b | 1634 | data_opt = EXT4_MOUNT_ORDERED_DATA; |
ac27a0ec DK |
1635 | goto datacheck; |
1636 | case Opt_data_writeback: | |
617ba13b | 1637 | data_opt = EXT4_MOUNT_WRITEBACK_DATA; |
ac27a0ec DK |
1638 | datacheck: |
1639 | if (is_remount) { | |
482a7425 | 1640 | if (test_opt(sb, DATA_FLAGS) != data_opt) { |
b31e1552 ES |
1641 | ext4_msg(sb, KERN_ERR, |
1642 | "Cannot change data mode on remount"); | |
ac27a0ec DK |
1643 | return 0; |
1644 | } | |
1645 | } else { | |
fd8c37ec | 1646 | clear_opt(sb, DATA_FLAGS); |
ac27a0ec DK |
1647 | sbi->s_mount_opt |= data_opt; |
1648 | } | |
1649 | break; | |
5bf5683a | 1650 | case Opt_data_err_abort: |
fd8c37ec | 1651 | set_opt(sb, DATA_ERR_ABORT); |
5bf5683a HK |
1652 | break; |
1653 | case Opt_data_err_ignore: | |
fd8c37ec | 1654 | clear_opt(sb, DATA_ERR_ABORT); |
5bf5683a | 1655 | break; |
ac27a0ec DK |
1656 | #ifdef CONFIG_QUOTA |
1657 | case Opt_usrjquota: | |
56c50f11 | 1658 | if (!set_qf_name(sb, USRQUOTA, &args[0])) |
ac27a0ec | 1659 | return 0; |
56c50f11 DM |
1660 | break; |
1661 | case Opt_grpjquota: | |
1662 | if (!set_qf_name(sb, GRPQUOTA, &args[0])) | |
ac27a0ec | 1663 | return 0; |
ac27a0ec DK |
1664 | break; |
1665 | case Opt_offusrjquota: | |
56c50f11 DM |
1666 | if (!clear_qf_name(sb, USRQUOTA)) |
1667 | return 0; | |
1668 | break; | |
ac27a0ec | 1669 | case Opt_offgrpjquota: |
56c50f11 | 1670 | if (!clear_qf_name(sb, GRPQUOTA)) |
ac27a0ec | 1671 | return 0; |
ac27a0ec | 1672 | break; |
56c50f11 | 1673 | |
ac27a0ec | 1674 | case Opt_jqfmt_vfsold: |
dfc5d03f JK |
1675 | qfmt = QFMT_VFS_OLD; |
1676 | goto set_qf_format; | |
ac27a0ec | 1677 | case Opt_jqfmt_vfsv0: |
dfc5d03f | 1678 | qfmt = QFMT_VFS_V0; |
5a20bdfc JK |
1679 | goto set_qf_format; |
1680 | case Opt_jqfmt_vfsv1: | |
1681 | qfmt = QFMT_VFS_V1; | |
dfc5d03f | 1682 | set_qf_format: |
17bd13b3 | 1683 | if (sb_any_quota_loaded(sb) && |
dfc5d03f | 1684 | sbi->s_jquota_fmt != qfmt) { |
b31e1552 | 1685 | ext4_msg(sb, KERN_ERR, "Cannot change " |
dfc5d03f | 1686 | "journaled quota options when " |
b31e1552 | 1687 | "quota turned on"); |
dfc5d03f JK |
1688 | return 0; |
1689 | } | |
1690 | sbi->s_jquota_fmt = qfmt; | |
ac27a0ec DK |
1691 | break; |
1692 | case Opt_quota: | |
1693 | case Opt_usrquota: | |
fd8c37ec TT |
1694 | set_opt(sb, QUOTA); |
1695 | set_opt(sb, USRQUOTA); | |
ac27a0ec DK |
1696 | break; |
1697 | case Opt_grpquota: | |
fd8c37ec TT |
1698 | set_opt(sb, QUOTA); |
1699 | set_opt(sb, GRPQUOTA); | |
ac27a0ec DK |
1700 | break; |
1701 | case Opt_noquota: | |
17bd13b3 | 1702 | if (sb_any_quota_loaded(sb)) { |
b31e1552 ES |
1703 | ext4_msg(sb, KERN_ERR, "Cannot change quota " |
1704 | "options when quota turned on"); | |
ac27a0ec DK |
1705 | return 0; |
1706 | } | |
fd8c37ec TT |
1707 | clear_opt(sb, QUOTA); |
1708 | clear_opt(sb, USRQUOTA); | |
1709 | clear_opt(sb, GRPQUOTA); | |
ac27a0ec DK |
1710 | break; |
1711 | #else | |
1712 | case Opt_quota: | |
1713 | case Opt_usrquota: | |
1714 | case Opt_grpquota: | |
b31e1552 ES |
1715 | ext4_msg(sb, KERN_ERR, |
1716 | "quota options not supported"); | |
cd59e7b9 | 1717 | break; |
ac27a0ec DK |
1718 | case Opt_usrjquota: |
1719 | case Opt_grpjquota: | |
1720 | case Opt_offusrjquota: | |
1721 | case Opt_offgrpjquota: | |
1722 | case Opt_jqfmt_vfsold: | |
1723 | case Opt_jqfmt_vfsv0: | |
5a20bdfc | 1724 | case Opt_jqfmt_vfsv1: |
b31e1552 ES |
1725 | ext4_msg(sb, KERN_ERR, |
1726 | "journaled quota options not supported"); | |
ac27a0ec DK |
1727 | break; |
1728 | case Opt_noquota: | |
1729 | break; | |
1730 | #endif | |
1731 | case Opt_abort: | |
4ab2f15b | 1732 | sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; |
ac27a0ec | 1733 | break; |
06705bff | 1734 | case Opt_nobarrier: |
fd8c37ec | 1735 | clear_opt(sb, BARRIER); |
06705bff | 1736 | break; |
ac27a0ec | 1737 | case Opt_barrier: |
15121c18 ES |
1738 | if (args[0].from) { |
1739 | if (match_int(&args[0], &option)) | |
1740 | return 0; | |
1741 | } else | |
1742 | option = 1; /* No argument, default to 1 */ | |
ac27a0ec | 1743 | if (option) |
fd8c37ec | 1744 | set_opt(sb, BARRIER); |
ac27a0ec | 1745 | else |
fd8c37ec | 1746 | clear_opt(sb, BARRIER); |
ac27a0ec DK |
1747 | break; |
1748 | case Opt_ignore: | |
1749 | break; | |
1750 | case Opt_resize: | |
1751 | if (!is_remount) { | |
b31e1552 ES |
1752 | ext4_msg(sb, KERN_ERR, |
1753 | "resize option only available " | |
1754 | "for remount"); | |
ac27a0ec DK |
1755 | return 0; |
1756 | } | |
1757 | if (match_int(&args[0], &option) != 0) | |
1758 | return 0; | |
1759 | *n_blocks_count = option; | |
1760 | break; | |
1761 | case Opt_nobh: | |
206f7ab4 CH |
1762 | ext4_msg(sb, KERN_WARNING, |
1763 | "Ignoring deprecated nobh option"); | |
ac27a0ec DK |
1764 | break; |
1765 | case Opt_bh: | |
206f7ab4 CH |
1766 | ext4_msg(sb, KERN_WARNING, |
1767 | "Ignoring deprecated bh option"); | |
ac27a0ec | 1768 | break; |
25ec56b5 | 1769 | case Opt_i_version: |
fd8c37ec | 1770 | set_opt(sb, I_VERSION); |
25ec56b5 JNC |
1771 | sb->s_flags |= MS_I_VERSION; |
1772 | break; | |
dd919b98 | 1773 | case Opt_nodelalloc: |
fd8c37ec | 1774 | clear_opt(sb, DELALLOC); |
dd919b98 | 1775 | break; |
1449032b | 1776 | case Opt_mblk_io_submit: |
fd8c37ec | 1777 | set_opt(sb, MBLK_IO_SUBMIT); |
1449032b TT |
1778 | break; |
1779 | case Opt_nomblk_io_submit: | |
fd8c37ec | 1780 | clear_opt(sb, MBLK_IO_SUBMIT); |
1449032b | 1781 | break; |
c9de560d AT |
1782 | case Opt_stripe: |
1783 | if (match_int(&args[0], &option)) | |
1784 | return 0; | |
1785 | if (option < 0) | |
1786 | return 0; | |
1787 | sbi->s_stripe = option; | |
1788 | break; | |
64769240 | 1789 | case Opt_delalloc: |
fd8c37ec | 1790 | set_opt(sb, DELALLOC); |
64769240 | 1791 | break; |
6fd058f7 | 1792 | case Opt_block_validity: |
fd8c37ec | 1793 | set_opt(sb, BLOCK_VALIDITY); |
6fd058f7 TT |
1794 | break; |
1795 | case Opt_noblock_validity: | |
fd8c37ec | 1796 | clear_opt(sb, BLOCK_VALIDITY); |
6fd058f7 | 1797 | break; |
240799cd TT |
1798 | case Opt_inode_readahead_blks: |
1799 | if (match_int(&args[0], &option)) | |
1800 | return 0; | |
1801 | if (option < 0 || option > (1 << 30)) | |
1802 | return 0; | |
5dbd571d | 1803 | if (option && !is_power_of_2(option)) { |
b31e1552 ES |
1804 | ext4_msg(sb, KERN_ERR, |
1805 | "EXT4-fs: inode_readahead_blks" | |
1806 | " must be a power of 2"); | |
3197ebdb TT |
1807 | return 0; |
1808 | } | |
240799cd TT |
1809 | sbi->s_inode_readahead_blks = option; |
1810 | break; | |
b3881f74 TT |
1811 | case Opt_journal_ioprio: |
1812 | if (match_int(&args[0], &option)) | |
1813 | return 0; | |
1814 | if (option < 0 || option > 7) | |
1815 | break; | |
1816 | *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, | |
1817 | option); | |
1818 | break; | |
06705bff | 1819 | case Opt_noauto_da_alloc: |
fd8c37ec | 1820 | set_opt(sb, NO_AUTO_DA_ALLOC); |
06705bff | 1821 | break; |
afd4672d | 1822 | case Opt_auto_da_alloc: |
15121c18 ES |
1823 | if (args[0].from) { |
1824 | if (match_int(&args[0], &option)) | |
1825 | return 0; | |
1826 | } else | |
1827 | option = 1; /* No argument, default to 1 */ | |
afd4672d | 1828 | if (option) |
fd8c37ec | 1829 | clear_opt(sb, NO_AUTO_DA_ALLOC); |
afd4672d | 1830 | else |
fd8c37ec | 1831 | set_opt(sb,NO_AUTO_DA_ALLOC); |
afd4672d | 1832 | break; |
5328e635 | 1833 | case Opt_discard: |
fd8c37ec | 1834 | set_opt(sb, DISCARD); |
5328e635 ES |
1835 | break; |
1836 | case Opt_nodiscard: | |
fd8c37ec | 1837 | clear_opt(sb, DISCARD); |
5328e635 | 1838 | break; |
744692dc | 1839 | case Opt_dioread_nolock: |
fd8c37ec | 1840 | set_opt(sb, DIOREAD_NOLOCK); |
744692dc JZ |
1841 | break; |
1842 | case Opt_dioread_lock: | |
fd8c37ec | 1843 | clear_opt(sb, DIOREAD_NOLOCK); |
744692dc | 1844 | break; |
bfff6873 | 1845 | case Opt_init_inode_table: |
fd8c37ec | 1846 | set_opt(sb, INIT_INODE_TABLE); |
bfff6873 LC |
1847 | if (args[0].from) { |
1848 | if (match_int(&args[0], &option)) | |
1849 | return 0; | |
1850 | } else | |
1851 | option = EXT4_DEF_LI_WAIT_MULT; | |
1852 | if (option < 0) | |
1853 | return 0; | |
1854 | sbi->s_li_wait_mult = option; | |
1855 | break; | |
1856 | case Opt_noinit_inode_table: | |
fd8c37ec | 1857 | clear_opt(sb, INIT_INODE_TABLE); |
bfff6873 | 1858 | break; |
ac27a0ec | 1859 | default: |
b31e1552 ES |
1860 | ext4_msg(sb, KERN_ERR, |
1861 | "Unrecognized mount option \"%s\" " | |
1862 | "or missing value", p); | |
ac27a0ec DK |
1863 | return 0; |
1864 | } | |
1865 | } | |
1866 | #ifdef CONFIG_QUOTA | |
1867 | if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { | |
482a7425 | 1868 | if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA]) |
fd8c37ec | 1869 | clear_opt(sb, USRQUOTA); |
ac27a0ec | 1870 | |
482a7425 | 1871 | if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA]) |
fd8c37ec | 1872 | clear_opt(sb, GRPQUOTA); |
ac27a0ec | 1873 | |
56c50f11 | 1874 | if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { |
b31e1552 ES |
1875 | ext4_msg(sb, KERN_ERR, "old and new quota " |
1876 | "format mixing"); | |
ac27a0ec DK |
1877 | return 0; |
1878 | } | |
1879 | ||
1880 | if (!sbi->s_jquota_fmt) { | |
b31e1552 ES |
1881 | ext4_msg(sb, KERN_ERR, "journaled quota format " |
1882 | "not specified"); | |
ac27a0ec DK |
1883 | return 0; |
1884 | } | |
1885 | } else { | |
1886 | if (sbi->s_jquota_fmt) { | |
b31e1552 | 1887 | ext4_msg(sb, KERN_ERR, "journaled quota format " |
2c8be6b2 | 1888 | "specified with no journaling " |
b31e1552 | 1889 | "enabled"); |
ac27a0ec DK |
1890 | return 0; |
1891 | } | |
1892 | } | |
1893 | #endif | |
1894 | return 1; | |
1895 | } | |
1896 | ||
617ba13b | 1897 | static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, |
ac27a0ec DK |
1898 | int read_only) |
1899 | { | |
617ba13b | 1900 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec DK |
1901 | int res = 0; |
1902 | ||
617ba13b | 1903 | if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { |
b31e1552 ES |
1904 | ext4_msg(sb, KERN_ERR, "revision level too high, " |
1905 | "forcing read-only mode"); | |
ac27a0ec DK |
1906 | res = MS_RDONLY; |
1907 | } | |
1908 | if (read_only) | |
1909 | return res; | |
617ba13b | 1910 | if (!(sbi->s_mount_state & EXT4_VALID_FS)) |
b31e1552 ES |
1911 | ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " |
1912 | "running e2fsck is recommended"); | |
617ba13b | 1913 | else if ((sbi->s_mount_state & EXT4_ERROR_FS)) |
b31e1552 ES |
1914 | ext4_msg(sb, KERN_WARNING, |
1915 | "warning: mounting fs with errors, " | |
1916 | "running e2fsck is recommended"); | |
ed3ce80a | 1917 | else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && |
ac27a0ec DK |
1918 | le16_to_cpu(es->s_mnt_count) >= |
1919 | (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) | |
b31e1552 ES |
1920 | ext4_msg(sb, KERN_WARNING, |
1921 | "warning: maximal mount count reached, " | |
1922 | "running e2fsck is recommended"); | |
ac27a0ec DK |
1923 | else if (le32_to_cpu(es->s_checkinterval) && |
1924 | (le32_to_cpu(es->s_lastcheck) + | |
1925 | le32_to_cpu(es->s_checkinterval) <= get_seconds())) | |
b31e1552 ES |
1926 | ext4_msg(sb, KERN_WARNING, |
1927 | "warning: checktime reached, " | |
1928 | "running e2fsck is recommended"); | |
0b8e58a1 | 1929 | if (!sbi->s_journal) |
0390131b | 1930 | es->s_state &= cpu_to_le16(~EXT4_VALID_FS); |
ac27a0ec | 1931 | if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) |
617ba13b | 1932 | es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); |
e8546d06 | 1933 | le16_add_cpu(&es->s_mnt_count, 1); |
ac27a0ec | 1934 | es->s_mtime = cpu_to_le32(get_seconds()); |
617ba13b | 1935 | ext4_update_dynamic_rev(sb); |
0390131b FM |
1936 | if (sbi->s_journal) |
1937 | EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
ac27a0ec | 1938 | |
e2d67052 | 1939 | ext4_commit_super(sb, 1); |
ac27a0ec | 1940 | if (test_opt(sb, DEBUG)) |
a9df9a49 | 1941 | printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " |
a2595b8a | 1942 | "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", |
ac27a0ec DK |
1943 | sb->s_blocksize, |
1944 | sbi->s_groups_count, | |
617ba13b MC |
1945 | EXT4_BLOCKS_PER_GROUP(sb), |
1946 | EXT4_INODES_PER_GROUP(sb), | |
a2595b8a | 1947 | sbi->s_mount_opt, sbi->s_mount_opt2); |
ac27a0ec | 1948 | |
ac27a0ec DK |
1949 | return res; |
1950 | } | |
1951 | ||
772cb7c8 JS |
1952 | static int ext4_fill_flex_info(struct super_block *sb) |
1953 | { | |
1954 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1955 | struct ext4_group_desc *gdp = NULL; | |
772cb7c8 JS |
1956 | ext4_group_t flex_group_count; |
1957 | ext4_group_t flex_group; | |
1958 | int groups_per_flex = 0; | |
c5ca7c76 | 1959 | size_t size; |
772cb7c8 JS |
1960 | int i; |
1961 | ||
503358ae TT |
1962 | sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; |
1963 | groups_per_flex = 1 << sbi->s_log_groups_per_flex; | |
1964 | ||
1965 | if (groups_per_flex < 2) { | |
772cb7c8 JS |
1966 | sbi->s_log_groups_per_flex = 0; |
1967 | return 1; | |
1968 | } | |
1969 | ||
c62a11fd FB |
1970 | /* We allocate both existing and potentially added groups */ |
1971 | flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) + | |
d94e99a6 AK |
1972 | ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) << |
1973 | EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex; | |
c5ca7c76 TT |
1974 | size = flex_group_count * sizeof(struct flex_groups); |
1975 | sbi->s_flex_groups = kzalloc(size, GFP_KERNEL); | |
1976 | if (sbi->s_flex_groups == NULL) { | |
94de56ab JP |
1977 | sbi->s_flex_groups = vzalloc(size); |
1978 | if (sbi->s_flex_groups == NULL) { | |
1979 | ext4_msg(sb, KERN_ERR, | |
1980 | "not enough memory for %u flex groups", | |
1981 | flex_group_count); | |
1982 | goto failed; | |
1983 | } | |
772cb7c8 | 1984 | } |
772cb7c8 | 1985 | |
772cb7c8 | 1986 | for (i = 0; i < sbi->s_groups_count; i++) { |
88b6edd1 | 1987 | gdp = ext4_get_group_desc(sb, i, NULL); |
772cb7c8 JS |
1988 | |
1989 | flex_group = ext4_flex_group(sbi, i); | |
7ad9bb65 TT |
1990 | atomic_add(ext4_free_inodes_count(sb, gdp), |
1991 | &sbi->s_flex_groups[flex_group].free_inodes); | |
1992 | atomic_add(ext4_free_blks_count(sb, gdp), | |
1993 | &sbi->s_flex_groups[flex_group].free_blocks); | |
1994 | atomic_add(ext4_used_dirs_count(sb, gdp), | |
1995 | &sbi->s_flex_groups[flex_group].used_dirs); | |
772cb7c8 JS |
1996 | } |
1997 | ||
1998 | return 1; | |
1999 | failed: | |
2000 | return 0; | |
2001 | } | |
2002 | ||
717d50e4 AD |
2003 | __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group, |
2004 | struct ext4_group_desc *gdp) | |
2005 | { | |
2006 | __u16 crc = 0; | |
2007 | ||
2008 | if (sbi->s_es->s_feature_ro_compat & | |
2009 | cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) { | |
2010 | int offset = offsetof(struct ext4_group_desc, bg_checksum); | |
2011 | __le32 le_group = cpu_to_le32(block_group); | |
2012 | ||
2013 | crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); | |
2014 | crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); | |
2015 | crc = crc16(crc, (__u8 *)gdp, offset); | |
2016 | offset += sizeof(gdp->bg_checksum); /* skip checksum */ | |
2017 | /* for checksum of struct ext4_group_desc do the rest...*/ | |
2018 | if ((sbi->s_es->s_feature_incompat & | |
2019 | cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) && | |
2020 | offset < le16_to_cpu(sbi->s_es->s_desc_size)) | |
2021 | crc = crc16(crc, (__u8 *)gdp + offset, | |
2022 | le16_to_cpu(sbi->s_es->s_desc_size) - | |
2023 | offset); | |
2024 | } | |
2025 | ||
2026 | return cpu_to_le16(crc); | |
2027 | } | |
2028 | ||
2029 | int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group, | |
2030 | struct ext4_group_desc *gdp) | |
2031 | { | |
2032 | if ((sbi->s_es->s_feature_ro_compat & | |
2033 | cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) && | |
2034 | (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp))) | |
2035 | return 0; | |
2036 | ||
2037 | return 1; | |
2038 | } | |
2039 | ||
ac27a0ec | 2040 | /* Called at mount-time, super-block is locked */ |
bfff6873 LC |
2041 | static int ext4_check_descriptors(struct super_block *sb, |
2042 | ext4_group_t *first_not_zeroed) | |
ac27a0ec | 2043 | { |
617ba13b MC |
2044 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
2045 | ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); | |
2046 | ext4_fsblk_t last_block; | |
bd81d8ee LV |
2047 | ext4_fsblk_t block_bitmap; |
2048 | ext4_fsblk_t inode_bitmap; | |
2049 | ext4_fsblk_t inode_table; | |
ce421581 | 2050 | int flexbg_flag = 0; |
bfff6873 | 2051 | ext4_group_t i, grp = sbi->s_groups_count; |
ac27a0ec | 2052 | |
ce421581 JS |
2053 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) |
2054 | flexbg_flag = 1; | |
2055 | ||
af5bc92d | 2056 | ext4_debug("Checking group descriptors"); |
ac27a0ec | 2057 | |
197cd65a AM |
2058 | for (i = 0; i < sbi->s_groups_count; i++) { |
2059 | struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); | |
2060 | ||
ce421581 | 2061 | if (i == sbi->s_groups_count - 1 || flexbg_flag) |
bd81d8ee | 2062 | last_block = ext4_blocks_count(sbi->s_es) - 1; |
ac27a0ec DK |
2063 | else |
2064 | last_block = first_block + | |
617ba13b | 2065 | (EXT4_BLOCKS_PER_GROUP(sb) - 1); |
ac27a0ec | 2066 | |
bfff6873 LC |
2067 | if ((grp == sbi->s_groups_count) && |
2068 | !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) | |
2069 | grp = i; | |
2070 | ||
8fadc143 | 2071 | block_bitmap = ext4_block_bitmap(sb, gdp); |
2b2d6d01 | 2072 | if (block_bitmap < first_block || block_bitmap > last_block) { |
b31e1552 | 2073 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 2074 | "Block bitmap for group %u not in group " |
b31e1552 | 2075 | "(block %llu)!", i, block_bitmap); |
ac27a0ec DK |
2076 | return 0; |
2077 | } | |
8fadc143 | 2078 | inode_bitmap = ext4_inode_bitmap(sb, gdp); |
2b2d6d01 | 2079 | if (inode_bitmap < first_block || inode_bitmap > last_block) { |
b31e1552 | 2080 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 2081 | "Inode bitmap for group %u not in group " |
b31e1552 | 2082 | "(block %llu)!", i, inode_bitmap); |
ac27a0ec DK |
2083 | return 0; |
2084 | } | |
8fadc143 | 2085 | inode_table = ext4_inode_table(sb, gdp); |
bd81d8ee | 2086 | if (inode_table < first_block || |
2b2d6d01 | 2087 | inode_table + sbi->s_itb_per_group - 1 > last_block) { |
b31e1552 | 2088 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 2089 | "Inode table for group %u not in group " |
b31e1552 | 2090 | "(block %llu)!", i, inode_table); |
ac27a0ec DK |
2091 | return 0; |
2092 | } | |
955ce5f5 | 2093 | ext4_lock_group(sb, i); |
717d50e4 | 2094 | if (!ext4_group_desc_csum_verify(sbi, i, gdp)) { |
b31e1552 ES |
2095 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
2096 | "Checksum for group %u failed (%u!=%u)", | |
2097 | i, le16_to_cpu(ext4_group_desc_csum(sbi, i, | |
2098 | gdp)), le16_to_cpu(gdp->bg_checksum)); | |
7ee1ec4c | 2099 | if (!(sb->s_flags & MS_RDONLY)) { |
955ce5f5 | 2100 | ext4_unlock_group(sb, i); |
8a266467 | 2101 | return 0; |
7ee1ec4c | 2102 | } |
717d50e4 | 2103 | } |
955ce5f5 | 2104 | ext4_unlock_group(sb, i); |
ce421581 JS |
2105 | if (!flexbg_flag) |
2106 | first_block += EXT4_BLOCKS_PER_GROUP(sb); | |
ac27a0ec | 2107 | } |
bfff6873 LC |
2108 | if (NULL != first_not_zeroed) |
2109 | *first_not_zeroed = grp; | |
ac27a0ec | 2110 | |
bd81d8ee | 2111 | ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb)); |
0b8e58a1 | 2112 | sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb)); |
ac27a0ec DK |
2113 | return 1; |
2114 | } | |
2115 | ||
617ba13b | 2116 | /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at |
ac27a0ec DK |
2117 | * the superblock) which were deleted from all directories, but held open by |
2118 | * a process at the time of a crash. We walk the list and try to delete these | |
2119 | * inodes at recovery time (only with a read-write filesystem). | |
2120 | * | |
2121 | * In order to keep the orphan inode chain consistent during traversal (in | |
2122 | * case of crash during recovery), we link each inode into the superblock | |
2123 | * orphan list_head and handle it the same way as an inode deletion during | |
2124 | * normal operation (which journals the operations for us). | |
2125 | * | |
2126 | * We only do an iget() and an iput() on each inode, which is very safe if we | |
2127 | * accidentally point at an in-use or already deleted inode. The worst that | |
2128 | * can happen in this case is that we get a "bit already cleared" message from | |
617ba13b | 2129 | * ext4_free_inode(). The only reason we would point at a wrong inode is if |
ac27a0ec DK |
2130 | * e2fsck was run on this filesystem, and it must have already done the orphan |
2131 | * inode cleanup for us, so we can safely abort without any further action. | |
2132 | */ | |
2b2d6d01 TT |
2133 | static void ext4_orphan_cleanup(struct super_block *sb, |
2134 | struct ext4_super_block *es) | |
ac27a0ec DK |
2135 | { |
2136 | unsigned int s_flags = sb->s_flags; | |
2137 | int nr_orphans = 0, nr_truncates = 0; | |
2138 | #ifdef CONFIG_QUOTA | |
2139 | int i; | |
2140 | #endif | |
2141 | if (!es->s_last_orphan) { | |
2142 | jbd_debug(4, "no orphan inodes to clean up\n"); | |
2143 | return; | |
2144 | } | |
2145 | ||
a8f48a95 | 2146 | if (bdev_read_only(sb->s_bdev)) { |
b31e1552 ES |
2147 | ext4_msg(sb, KERN_ERR, "write access " |
2148 | "unavailable, skipping orphan cleanup"); | |
a8f48a95 ES |
2149 | return; |
2150 | } | |
2151 | ||
d39195c3 AG |
2152 | /* Check if feature set would not allow a r/w mount */ |
2153 | if (!ext4_feature_set_ok(sb, 0)) { | |
2154 | ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " | |
2155 | "unknown ROCOMPAT features"); | |
2156 | return; | |
2157 | } | |
2158 | ||
617ba13b | 2159 | if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { |
ac27a0ec DK |
2160 | if (es->s_last_orphan) |
2161 | jbd_debug(1, "Errors on filesystem, " | |
2162 | "clearing orphan list.\n"); | |
2163 | es->s_last_orphan = 0; | |
2164 | jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); | |
2165 | return; | |
2166 | } | |
2167 | ||
2168 | if (s_flags & MS_RDONLY) { | |
b31e1552 | 2169 | ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); |
ac27a0ec DK |
2170 | sb->s_flags &= ~MS_RDONLY; |
2171 | } | |
2172 | #ifdef CONFIG_QUOTA | |
2173 | /* Needed for iput() to work correctly and not trash data */ | |
2174 | sb->s_flags |= MS_ACTIVE; | |
2175 | /* Turn on quotas so that they are updated correctly */ | |
2176 | for (i = 0; i < MAXQUOTAS; i++) { | |
617ba13b MC |
2177 | if (EXT4_SB(sb)->s_qf_names[i]) { |
2178 | int ret = ext4_quota_on_mount(sb, i); | |
ac27a0ec | 2179 | if (ret < 0) |
b31e1552 ES |
2180 | ext4_msg(sb, KERN_ERR, |
2181 | "Cannot turn on journaled " | |
2182 | "quota: error %d", ret); | |
ac27a0ec DK |
2183 | } |
2184 | } | |
2185 | #endif | |
2186 | ||
2187 | while (es->s_last_orphan) { | |
2188 | struct inode *inode; | |
2189 | ||
97bd42b9 JB |
2190 | inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); |
2191 | if (IS_ERR(inode)) { | |
ac27a0ec DK |
2192 | es->s_last_orphan = 0; |
2193 | break; | |
2194 | } | |
2195 | ||
617ba13b | 2196 | list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); |
871a2931 | 2197 | dquot_initialize(inode); |
ac27a0ec | 2198 | if (inode->i_nlink) { |
b31e1552 ES |
2199 | ext4_msg(sb, KERN_DEBUG, |
2200 | "%s: truncating inode %lu to %lld bytes", | |
46e665e9 | 2201 | __func__, inode->i_ino, inode->i_size); |
e5f8eab8 | 2202 | jbd_debug(2, "truncating inode %lu to %lld bytes\n", |
ac27a0ec | 2203 | inode->i_ino, inode->i_size); |
617ba13b | 2204 | ext4_truncate(inode); |
ac27a0ec DK |
2205 | nr_truncates++; |
2206 | } else { | |
b31e1552 ES |
2207 | ext4_msg(sb, KERN_DEBUG, |
2208 | "%s: deleting unreferenced inode %lu", | |
46e665e9 | 2209 | __func__, inode->i_ino); |
ac27a0ec DK |
2210 | jbd_debug(2, "deleting unreferenced inode %lu\n", |
2211 | inode->i_ino); | |
2212 | nr_orphans++; | |
2213 | } | |
2214 | iput(inode); /* The delete magic happens here! */ | |
2215 | } | |
2216 | ||
2b2d6d01 | 2217 | #define PLURAL(x) (x), ((x) == 1) ? "" : "s" |
ac27a0ec DK |
2218 | |
2219 | if (nr_orphans) | |
b31e1552 ES |
2220 | ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", |
2221 | PLURAL(nr_orphans)); | |
ac27a0ec | 2222 | if (nr_truncates) |
b31e1552 ES |
2223 | ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", |
2224 | PLURAL(nr_truncates)); | |
ac27a0ec DK |
2225 | #ifdef CONFIG_QUOTA |
2226 | /* Turn quotas off */ | |
2227 | for (i = 0; i < MAXQUOTAS; i++) { | |
2228 | if (sb_dqopt(sb)->files[i]) | |
287a8095 | 2229 | dquot_quota_off(sb, i); |
ac27a0ec DK |
2230 | } |
2231 | #endif | |
2232 | sb->s_flags = s_flags; /* Restore MS_RDONLY status */ | |
2233 | } | |
0b8e58a1 | 2234 | |
cd2291a4 ES |
2235 | /* |
2236 | * Maximal extent format file size. | |
2237 | * Resulting logical blkno at s_maxbytes must fit in our on-disk | |
2238 | * extent format containers, within a sector_t, and within i_blocks | |
2239 | * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, | |
2240 | * so that won't be a limiting factor. | |
2241 | * | |
2242 | * Note, this does *not* consider any metadata overhead for vfs i_blocks. | |
2243 | */ | |
f287a1a5 | 2244 | static loff_t ext4_max_size(int blkbits, int has_huge_files) |
cd2291a4 ES |
2245 | { |
2246 | loff_t res; | |
2247 | loff_t upper_limit = MAX_LFS_FILESIZE; | |
2248 | ||
2249 | /* small i_blocks in vfs inode? */ | |
f287a1a5 | 2250 | if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { |
cd2291a4 | 2251 | /* |
90c699a9 | 2252 | * CONFIG_LBDAF is not enabled implies the inode |
cd2291a4 ES |
2253 | * i_block represent total blocks in 512 bytes |
2254 | * 32 == size of vfs inode i_blocks * 8 | |
2255 | */ | |
2256 | upper_limit = (1LL << 32) - 1; | |
2257 | ||
2258 | /* total blocks in file system block size */ | |
2259 | upper_limit >>= (blkbits - 9); | |
2260 | upper_limit <<= blkbits; | |
2261 | } | |
2262 | ||
2263 | /* 32-bit extent-start container, ee_block */ | |
2264 | res = 1LL << 32; | |
2265 | res <<= blkbits; | |
2266 | res -= 1; | |
2267 | ||
2268 | /* Sanity check against vm- & vfs- imposed limits */ | |
2269 | if (res > upper_limit) | |
2270 | res = upper_limit; | |
2271 | ||
2272 | return res; | |
2273 | } | |
ac27a0ec | 2274 | |
ac27a0ec | 2275 | /* |
cd2291a4 | 2276 | * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect |
0fc1b451 AK |
2277 | * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. |
2278 | * We need to be 1 filesystem block less than the 2^48 sector limit. | |
ac27a0ec | 2279 | */ |
f287a1a5 | 2280 | static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) |
ac27a0ec | 2281 | { |
617ba13b | 2282 | loff_t res = EXT4_NDIR_BLOCKS; |
0fc1b451 AK |
2283 | int meta_blocks; |
2284 | loff_t upper_limit; | |
0b8e58a1 AD |
2285 | /* This is calculated to be the largest file size for a dense, block |
2286 | * mapped file such that the file's total number of 512-byte sectors, | |
2287 | * including data and all indirect blocks, does not exceed (2^48 - 1). | |
2288 | * | |
2289 | * __u32 i_blocks_lo and _u16 i_blocks_high represent the total | |
2290 | * number of 512-byte sectors of the file. | |
0fc1b451 AK |
2291 | */ |
2292 | ||
f287a1a5 | 2293 | if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { |
0fc1b451 | 2294 | /* |
90c699a9 | 2295 | * !has_huge_files or CONFIG_LBDAF not enabled implies that |
0b8e58a1 AD |
2296 | * the inode i_block field represents total file blocks in |
2297 | * 2^32 512-byte sectors == size of vfs inode i_blocks * 8 | |
0fc1b451 AK |
2298 | */ |
2299 | upper_limit = (1LL << 32) - 1; | |
2300 | ||
2301 | /* total blocks in file system block size */ | |
2302 | upper_limit >>= (bits - 9); | |
2303 | ||
2304 | } else { | |
8180a562 AK |
2305 | /* |
2306 | * We use 48 bit ext4_inode i_blocks | |
2307 | * With EXT4_HUGE_FILE_FL set the i_blocks | |
2308 | * represent total number of blocks in | |
2309 | * file system block size | |
2310 | */ | |
0fc1b451 AK |
2311 | upper_limit = (1LL << 48) - 1; |
2312 | ||
0fc1b451 AK |
2313 | } |
2314 | ||
2315 | /* indirect blocks */ | |
2316 | meta_blocks = 1; | |
2317 | /* double indirect blocks */ | |
2318 | meta_blocks += 1 + (1LL << (bits-2)); | |
2319 | /* tripple indirect blocks */ | |
2320 | meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); | |
2321 | ||
2322 | upper_limit -= meta_blocks; | |
2323 | upper_limit <<= bits; | |
ac27a0ec DK |
2324 | |
2325 | res += 1LL << (bits-2); | |
2326 | res += 1LL << (2*(bits-2)); | |
2327 | res += 1LL << (3*(bits-2)); | |
2328 | res <<= bits; | |
2329 | if (res > upper_limit) | |
2330 | res = upper_limit; | |
0fc1b451 AK |
2331 | |
2332 | if (res > MAX_LFS_FILESIZE) | |
2333 | res = MAX_LFS_FILESIZE; | |
2334 | ||
ac27a0ec DK |
2335 | return res; |
2336 | } | |
2337 | ||
617ba13b | 2338 | static ext4_fsblk_t descriptor_loc(struct super_block *sb, |
0b8e58a1 | 2339 | ext4_fsblk_t logical_sb_block, int nr) |
ac27a0ec | 2340 | { |
617ba13b | 2341 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
fd2d4291 | 2342 | ext4_group_t bg, first_meta_bg; |
ac27a0ec DK |
2343 | int has_super = 0; |
2344 | ||
2345 | first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); | |
2346 | ||
617ba13b | 2347 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) || |
ac27a0ec | 2348 | nr < first_meta_bg) |
70bbb3e0 | 2349 | return logical_sb_block + nr + 1; |
ac27a0ec | 2350 | bg = sbi->s_desc_per_block * nr; |
617ba13b | 2351 | if (ext4_bg_has_super(sb, bg)) |
ac27a0ec | 2352 | has_super = 1; |
0b8e58a1 | 2353 | |
617ba13b | 2354 | return (has_super + ext4_group_first_block_no(sb, bg)); |
ac27a0ec DK |
2355 | } |
2356 | ||
c9de560d AT |
2357 | /** |
2358 | * ext4_get_stripe_size: Get the stripe size. | |
2359 | * @sbi: In memory super block info | |
2360 | * | |
2361 | * If we have specified it via mount option, then | |
2362 | * use the mount option value. If the value specified at mount time is | |
2363 | * greater than the blocks per group use the super block value. | |
2364 | * If the super block value is greater than blocks per group return 0. | |
2365 | * Allocator needs it be less than blocks per group. | |
2366 | * | |
2367 | */ | |
2368 | static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) | |
2369 | { | |
2370 | unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); | |
2371 | unsigned long stripe_width = | |
2372 | le32_to_cpu(sbi->s_es->s_raid_stripe_width); | |
2373 | ||
2374 | if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) | |
2375 | return sbi->s_stripe; | |
2376 | ||
2377 | if (stripe_width <= sbi->s_blocks_per_group) | |
2378 | return stripe_width; | |
2379 | ||
2380 | if (stride <= sbi->s_blocks_per_group) | |
2381 | return stride; | |
2382 | ||
2383 | return 0; | |
2384 | } | |
ac27a0ec | 2385 | |
3197ebdb TT |
2386 | /* sysfs supprt */ |
2387 | ||
2388 | struct ext4_attr { | |
2389 | struct attribute attr; | |
2390 | ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *); | |
60e6679e | 2391 | ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *, |
3197ebdb TT |
2392 | const char *, size_t); |
2393 | int offset; | |
2394 | }; | |
2395 | ||
2396 | static int parse_strtoul(const char *buf, | |
2397 | unsigned long max, unsigned long *value) | |
2398 | { | |
2399 | char *endp; | |
2400 | ||
e7d2860b AGR |
2401 | *value = simple_strtoul(skip_spaces(buf), &endp, 0); |
2402 | endp = skip_spaces(endp); | |
3197ebdb TT |
2403 | if (*endp || *value > max) |
2404 | return -EINVAL; | |
2405 | ||
2406 | return 0; | |
2407 | } | |
2408 | ||
2409 | static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a, | |
2410 | struct ext4_sb_info *sbi, | |
2411 | char *buf) | |
2412 | { | |
2413 | return snprintf(buf, PAGE_SIZE, "%llu\n", | |
2414 | (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter)); | |
2415 | } | |
2416 | ||
2417 | static ssize_t session_write_kbytes_show(struct ext4_attr *a, | |
2418 | struct ext4_sb_info *sbi, char *buf) | |
2419 | { | |
2420 | struct super_block *sb = sbi->s_buddy_cache->i_sb; | |
2421 | ||
f613dfcb TT |
2422 | if (!sb->s_bdev->bd_part) |
2423 | return snprintf(buf, PAGE_SIZE, "0\n"); | |
3197ebdb TT |
2424 | return snprintf(buf, PAGE_SIZE, "%lu\n", |
2425 | (part_stat_read(sb->s_bdev->bd_part, sectors[1]) - | |
2426 | sbi->s_sectors_written_start) >> 1); | |
2427 | } | |
2428 | ||
2429 | static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a, | |
2430 | struct ext4_sb_info *sbi, char *buf) | |
2431 | { | |
2432 | struct super_block *sb = sbi->s_buddy_cache->i_sb; | |
2433 | ||
f613dfcb TT |
2434 | if (!sb->s_bdev->bd_part) |
2435 | return snprintf(buf, PAGE_SIZE, "0\n"); | |
3197ebdb | 2436 | return snprintf(buf, PAGE_SIZE, "%llu\n", |
a6b43e38 | 2437 | (unsigned long long)(sbi->s_kbytes_written + |
3197ebdb | 2438 | ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - |
a6b43e38 | 2439 | EXT4_SB(sb)->s_sectors_written_start) >> 1))); |
3197ebdb TT |
2440 | } |
2441 | ||
77f4135f VH |
2442 | static ssize_t extent_cache_hits_show(struct ext4_attr *a, |
2443 | struct ext4_sb_info *sbi, char *buf) | |
2444 | { | |
2445 | return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits); | |
2446 | } | |
2447 | ||
2448 | static ssize_t extent_cache_misses_show(struct ext4_attr *a, | |
2449 | struct ext4_sb_info *sbi, char *buf) | |
2450 | { | |
2451 | return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses); | |
2452 | } | |
2453 | ||
3197ebdb TT |
2454 | static ssize_t inode_readahead_blks_store(struct ext4_attr *a, |
2455 | struct ext4_sb_info *sbi, | |
2456 | const char *buf, size_t count) | |
2457 | { | |
2458 | unsigned long t; | |
2459 | ||
2460 | if (parse_strtoul(buf, 0x40000000, &t)) | |
2461 | return -EINVAL; | |
2462 | ||
5dbd571d | 2463 | if (t && !is_power_of_2(t)) |
3197ebdb TT |
2464 | return -EINVAL; |
2465 | ||
2466 | sbi->s_inode_readahead_blks = t; | |
2467 | return count; | |
2468 | } | |
2469 | ||
2470 | static ssize_t sbi_ui_show(struct ext4_attr *a, | |
0b8e58a1 | 2471 | struct ext4_sb_info *sbi, char *buf) |
3197ebdb TT |
2472 | { |
2473 | unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset); | |
2474 | ||
2475 | return snprintf(buf, PAGE_SIZE, "%u\n", *ui); | |
2476 | } | |
2477 | ||
2478 | static ssize_t sbi_ui_store(struct ext4_attr *a, | |
2479 | struct ext4_sb_info *sbi, | |
2480 | const char *buf, size_t count) | |
2481 | { | |
2482 | unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset); | |
2483 | unsigned long t; | |
2484 | ||
2485 | if (parse_strtoul(buf, 0xffffffff, &t)) | |
2486 | return -EINVAL; | |
2487 | *ui = t; | |
2488 | return count; | |
2489 | } | |
2490 | ||
2491 | #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \ | |
2492 | static struct ext4_attr ext4_attr_##_name = { \ | |
2493 | .attr = {.name = __stringify(_name), .mode = _mode }, \ | |
2494 | .show = _show, \ | |
2495 | .store = _store, \ | |
2496 | .offset = offsetof(struct ext4_sb_info, _elname), \ | |
2497 | } | |
2498 | #define EXT4_ATTR(name, mode, show, store) \ | |
2499 | static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store) | |
2500 | ||
857ac889 | 2501 | #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL) |
3197ebdb TT |
2502 | #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL) |
2503 | #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store) | |
2504 | #define EXT4_RW_ATTR_SBI_UI(name, elname) \ | |
2505 | EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname) | |
2506 | #define ATTR_LIST(name) &ext4_attr_##name.attr | |
2507 | ||
2508 | EXT4_RO_ATTR(delayed_allocation_blocks); | |
2509 | EXT4_RO_ATTR(session_write_kbytes); | |
2510 | EXT4_RO_ATTR(lifetime_write_kbytes); | |
77f4135f VH |
2511 | EXT4_RO_ATTR(extent_cache_hits); |
2512 | EXT4_RO_ATTR(extent_cache_misses); | |
3197ebdb TT |
2513 | EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show, |
2514 | inode_readahead_blks_store, s_inode_readahead_blks); | |
11013911 | 2515 | EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal); |
3197ebdb TT |
2516 | EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats); |
2517 | EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan); | |
2518 | EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan); | |
2519 | EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs); | |
2520 | EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request); | |
2521 | EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc); | |
55138e0b | 2522 | EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump); |
3197ebdb TT |
2523 | |
2524 | static struct attribute *ext4_attrs[] = { | |
2525 | ATTR_LIST(delayed_allocation_blocks), | |
2526 | ATTR_LIST(session_write_kbytes), | |
2527 | ATTR_LIST(lifetime_write_kbytes), | |
77f4135f VH |
2528 | ATTR_LIST(extent_cache_hits), |
2529 | ATTR_LIST(extent_cache_misses), | |
3197ebdb | 2530 | ATTR_LIST(inode_readahead_blks), |
11013911 | 2531 | ATTR_LIST(inode_goal), |
3197ebdb TT |
2532 | ATTR_LIST(mb_stats), |
2533 | ATTR_LIST(mb_max_to_scan), | |
2534 | ATTR_LIST(mb_min_to_scan), | |
2535 | ATTR_LIST(mb_order2_req), | |
2536 | ATTR_LIST(mb_stream_req), | |
2537 | ATTR_LIST(mb_group_prealloc), | |
55138e0b | 2538 | ATTR_LIST(max_writeback_mb_bump), |
3197ebdb TT |
2539 | NULL, |
2540 | }; | |
2541 | ||
857ac889 LC |
2542 | /* Features this copy of ext4 supports */ |
2543 | EXT4_INFO_ATTR(lazy_itable_init); | |
27ee40df | 2544 | EXT4_INFO_ATTR(batched_discard); |
857ac889 LC |
2545 | |
2546 | static struct attribute *ext4_feat_attrs[] = { | |
2547 | ATTR_LIST(lazy_itable_init), | |
27ee40df | 2548 | ATTR_LIST(batched_discard), |
857ac889 LC |
2549 | NULL, |
2550 | }; | |
2551 | ||
3197ebdb TT |
2552 | static ssize_t ext4_attr_show(struct kobject *kobj, |
2553 | struct attribute *attr, char *buf) | |
2554 | { | |
2555 | struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, | |
2556 | s_kobj); | |
2557 | struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); | |
2558 | ||
2559 | return a->show ? a->show(a, sbi, buf) : 0; | |
2560 | } | |
2561 | ||
2562 | static ssize_t ext4_attr_store(struct kobject *kobj, | |
2563 | struct attribute *attr, | |
2564 | const char *buf, size_t len) | |
2565 | { | |
2566 | struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, | |
2567 | s_kobj); | |
2568 | struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); | |
2569 | ||
2570 | return a->store ? a->store(a, sbi, buf, len) : 0; | |
2571 | } | |
2572 | ||
2573 | static void ext4_sb_release(struct kobject *kobj) | |
2574 | { | |
2575 | struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, | |
2576 | s_kobj); | |
2577 | complete(&sbi->s_kobj_unregister); | |
2578 | } | |
2579 | ||
52cf25d0 | 2580 | static const struct sysfs_ops ext4_attr_ops = { |
3197ebdb TT |
2581 | .show = ext4_attr_show, |
2582 | .store = ext4_attr_store, | |
2583 | }; | |
2584 | ||
2585 | static struct kobj_type ext4_ktype = { | |
2586 | .default_attrs = ext4_attrs, | |
2587 | .sysfs_ops = &ext4_attr_ops, | |
2588 | .release = ext4_sb_release, | |
2589 | }; | |
2590 | ||
857ac889 LC |
2591 | static void ext4_feat_release(struct kobject *kobj) |
2592 | { | |
2593 | complete(&ext4_feat->f_kobj_unregister); | |
2594 | } | |
2595 | ||
2596 | static struct kobj_type ext4_feat_ktype = { | |
2597 | .default_attrs = ext4_feat_attrs, | |
2598 | .sysfs_ops = &ext4_attr_ops, | |
2599 | .release = ext4_feat_release, | |
2600 | }; | |
2601 | ||
a13fb1a4 ES |
2602 | /* |
2603 | * Check whether this filesystem can be mounted based on | |
2604 | * the features present and the RDONLY/RDWR mount requested. | |
2605 | * Returns 1 if this filesystem can be mounted as requested, | |
2606 | * 0 if it cannot be. | |
2607 | */ | |
2608 | static int ext4_feature_set_ok(struct super_block *sb, int readonly) | |
2609 | { | |
2610 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) { | |
2611 | ext4_msg(sb, KERN_ERR, | |
2612 | "Couldn't mount because of " | |
2613 | "unsupported optional features (%x)", | |
2614 | (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & | |
2615 | ~EXT4_FEATURE_INCOMPAT_SUPP)); | |
2616 | return 0; | |
2617 | } | |
2618 | ||
2619 | if (readonly) | |
2620 | return 1; | |
2621 | ||
2622 | /* Check that feature set is OK for a read-write mount */ | |
2623 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) { | |
2624 | ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " | |
2625 | "unsupported optional features (%x)", | |
2626 | (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & | |
2627 | ~EXT4_FEATURE_RO_COMPAT_SUPP)); | |
2628 | return 0; | |
2629 | } | |
2630 | /* | |
2631 | * Large file size enabled file system can only be mounted | |
2632 | * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF | |
2633 | */ | |
2634 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) { | |
2635 | if (sizeof(blkcnt_t) < sizeof(u64)) { | |
2636 | ext4_msg(sb, KERN_ERR, "Filesystem with huge files " | |
2637 | "cannot be mounted RDWR without " | |
2638 | "CONFIG_LBDAF"); | |
2639 | return 0; | |
2640 | } | |
2641 | } | |
2642 | return 1; | |
2643 | } | |
2644 | ||
66e61a9e TT |
2645 | /* |
2646 | * This function is called once a day if we have errors logged | |
2647 | * on the file system | |
2648 | */ | |
2649 | static void print_daily_error_info(unsigned long arg) | |
2650 | { | |
2651 | struct super_block *sb = (struct super_block *) arg; | |
2652 | struct ext4_sb_info *sbi; | |
2653 | struct ext4_super_block *es; | |
2654 | ||
2655 | sbi = EXT4_SB(sb); | |
2656 | es = sbi->s_es; | |
2657 | ||
2658 | if (es->s_error_count) | |
2659 | ext4_msg(sb, KERN_NOTICE, "error count: %u", | |
2660 | le32_to_cpu(es->s_error_count)); | |
2661 | if (es->s_first_error_time) { | |
2662 | printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d", | |
2663 | sb->s_id, le32_to_cpu(es->s_first_error_time), | |
2664 | (int) sizeof(es->s_first_error_func), | |
2665 | es->s_first_error_func, | |
2666 | le32_to_cpu(es->s_first_error_line)); | |
2667 | if (es->s_first_error_ino) | |
2668 | printk(": inode %u", | |
2669 | le32_to_cpu(es->s_first_error_ino)); | |
2670 | if (es->s_first_error_block) | |
2671 | printk(": block %llu", (unsigned long long) | |
2672 | le64_to_cpu(es->s_first_error_block)); | |
2673 | printk("\n"); | |
2674 | } | |
2675 | if (es->s_last_error_time) { | |
2676 | printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d", | |
2677 | sb->s_id, le32_to_cpu(es->s_last_error_time), | |
2678 | (int) sizeof(es->s_last_error_func), | |
2679 | es->s_last_error_func, | |
2680 | le32_to_cpu(es->s_last_error_line)); | |
2681 | if (es->s_last_error_ino) | |
2682 | printk(": inode %u", | |
2683 | le32_to_cpu(es->s_last_error_ino)); | |
2684 | if (es->s_last_error_block) | |
2685 | printk(": block %llu", (unsigned long long) | |
2686 | le64_to_cpu(es->s_last_error_block)); | |
2687 | printk("\n"); | |
2688 | } | |
2689 | mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ | |
2690 | } | |
2691 | ||
bfff6873 LC |
2692 | /* Find next suitable group and run ext4_init_inode_table */ |
2693 | static int ext4_run_li_request(struct ext4_li_request *elr) | |
2694 | { | |
2695 | struct ext4_group_desc *gdp = NULL; | |
2696 | ext4_group_t group, ngroups; | |
2697 | struct super_block *sb; | |
2698 | unsigned long timeout = 0; | |
2699 | int ret = 0; | |
2700 | ||
2701 | sb = elr->lr_super; | |
2702 | ngroups = EXT4_SB(sb)->s_groups_count; | |
2703 | ||
2704 | for (group = elr->lr_next_group; group < ngroups; group++) { | |
2705 | gdp = ext4_get_group_desc(sb, group, NULL); | |
2706 | if (!gdp) { | |
2707 | ret = 1; | |
2708 | break; | |
2709 | } | |
2710 | ||
2711 | if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) | |
2712 | break; | |
2713 | } | |
2714 | ||
2715 | if (group == ngroups) | |
2716 | ret = 1; | |
2717 | ||
2718 | if (!ret) { | |
2719 | timeout = jiffies; | |
2720 | ret = ext4_init_inode_table(sb, group, | |
2721 | elr->lr_timeout ? 0 : 1); | |
2722 | if (elr->lr_timeout == 0) { | |
51ce6511 LC |
2723 | timeout = (jiffies - timeout) * |
2724 | elr->lr_sbi->s_li_wait_mult; | |
bfff6873 LC |
2725 | elr->lr_timeout = timeout; |
2726 | } | |
2727 | elr->lr_next_sched = jiffies + elr->lr_timeout; | |
2728 | elr->lr_next_group = group + 1; | |
2729 | } | |
2730 | ||
2731 | return ret; | |
2732 | } | |
2733 | ||
2734 | /* | |
2735 | * Remove lr_request from the list_request and free the | |
4ed5c033 | 2736 | * request structure. Should be called with li_list_mtx held |
bfff6873 LC |
2737 | */ |
2738 | static void ext4_remove_li_request(struct ext4_li_request *elr) | |
2739 | { | |
2740 | struct ext4_sb_info *sbi; | |
2741 | ||
2742 | if (!elr) | |
2743 | return; | |
2744 | ||
2745 | sbi = elr->lr_sbi; | |
2746 | ||
2747 | list_del(&elr->lr_request); | |
2748 | sbi->s_li_request = NULL; | |
2749 | kfree(elr); | |
2750 | } | |
2751 | ||
2752 | static void ext4_unregister_li_request(struct super_block *sb) | |
2753 | { | |
1bb933fb LC |
2754 | mutex_lock(&ext4_li_mtx); |
2755 | if (!ext4_li_info) { | |
2756 | mutex_unlock(&ext4_li_mtx); | |
bfff6873 | 2757 | return; |
1bb933fb | 2758 | } |
bfff6873 LC |
2759 | |
2760 | mutex_lock(&ext4_li_info->li_list_mtx); | |
1bb933fb | 2761 | ext4_remove_li_request(EXT4_SB(sb)->s_li_request); |
bfff6873 | 2762 | mutex_unlock(&ext4_li_info->li_list_mtx); |
1bb933fb | 2763 | mutex_unlock(&ext4_li_mtx); |
bfff6873 LC |
2764 | } |
2765 | ||
8f1f7453 ES |
2766 | static struct task_struct *ext4_lazyinit_task; |
2767 | ||
bfff6873 LC |
2768 | /* |
2769 | * This is the function where ext4lazyinit thread lives. It walks | |
2770 | * through the request list searching for next scheduled filesystem. | |
2771 | * When such a fs is found, run the lazy initialization request | |
2772 | * (ext4_rn_li_request) and keep track of the time spend in this | |
2773 | * function. Based on that time we compute next schedule time of | |
2774 | * the request. When walking through the list is complete, compute | |
2775 | * next waking time and put itself into sleep. | |
2776 | */ | |
2777 | static int ext4_lazyinit_thread(void *arg) | |
2778 | { | |
2779 | struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; | |
2780 | struct list_head *pos, *n; | |
2781 | struct ext4_li_request *elr; | |
4ed5c033 | 2782 | unsigned long next_wakeup, cur; |
bfff6873 LC |
2783 | |
2784 | BUG_ON(NULL == eli); | |
2785 | ||
bfff6873 LC |
2786 | cont_thread: |
2787 | while (true) { | |
2788 | next_wakeup = MAX_JIFFY_OFFSET; | |
2789 | ||
2790 | mutex_lock(&eli->li_list_mtx); | |
2791 | if (list_empty(&eli->li_request_list)) { | |
2792 | mutex_unlock(&eli->li_list_mtx); | |
2793 | goto exit_thread; | |
2794 | } | |
2795 | ||
2796 | list_for_each_safe(pos, n, &eli->li_request_list) { | |
2797 | elr = list_entry(pos, struct ext4_li_request, | |
2798 | lr_request); | |
2799 | ||
b2c78cd0 TT |
2800 | if (time_after_eq(jiffies, elr->lr_next_sched)) { |
2801 | if (ext4_run_li_request(elr) != 0) { | |
2802 | /* error, remove the lazy_init job */ | |
2803 | ext4_remove_li_request(elr); | |
2804 | continue; | |
2805 | } | |
bfff6873 LC |
2806 | } |
2807 | ||
2808 | if (time_before(elr->lr_next_sched, next_wakeup)) | |
2809 | next_wakeup = elr->lr_next_sched; | |
2810 | } | |
2811 | mutex_unlock(&eli->li_list_mtx); | |
2812 | ||
2813 | if (freezing(current)) | |
2814 | refrigerator(); | |
2815 | ||
4ed5c033 LC |
2816 | cur = jiffies; |
2817 | if ((time_after_eq(cur, next_wakeup)) || | |
f4245bd4 | 2818 | (MAX_JIFFY_OFFSET == next_wakeup)) { |
bfff6873 LC |
2819 | cond_resched(); |
2820 | continue; | |
2821 | } | |
2822 | ||
4ed5c033 LC |
2823 | schedule_timeout_interruptible(next_wakeup - cur); |
2824 | ||
8f1f7453 ES |
2825 | if (kthread_should_stop()) { |
2826 | ext4_clear_request_list(); | |
2827 | goto exit_thread; | |
2828 | } | |
bfff6873 LC |
2829 | } |
2830 | ||
2831 | exit_thread: | |
2832 | /* | |
2833 | * It looks like the request list is empty, but we need | |
2834 | * to check it under the li_list_mtx lock, to prevent any | |
2835 | * additions into it, and of course we should lock ext4_li_mtx | |
2836 | * to atomically free the list and ext4_li_info, because at | |
2837 | * this point another ext4 filesystem could be registering | |
2838 | * new one. | |
2839 | */ | |
2840 | mutex_lock(&ext4_li_mtx); | |
2841 | mutex_lock(&eli->li_list_mtx); | |
2842 | if (!list_empty(&eli->li_request_list)) { | |
2843 | mutex_unlock(&eli->li_list_mtx); | |
2844 | mutex_unlock(&ext4_li_mtx); | |
2845 | goto cont_thread; | |
2846 | } | |
2847 | mutex_unlock(&eli->li_list_mtx); | |
bfff6873 LC |
2848 | kfree(ext4_li_info); |
2849 | ext4_li_info = NULL; | |
2850 | mutex_unlock(&ext4_li_mtx); | |
2851 | ||
2852 | return 0; | |
2853 | } | |
2854 | ||
2855 | static void ext4_clear_request_list(void) | |
2856 | { | |
2857 | struct list_head *pos, *n; | |
2858 | struct ext4_li_request *elr; | |
2859 | ||
2860 | mutex_lock(&ext4_li_info->li_list_mtx); | |
bfff6873 LC |
2861 | list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { |
2862 | elr = list_entry(pos, struct ext4_li_request, | |
2863 | lr_request); | |
2864 | ext4_remove_li_request(elr); | |
2865 | } | |
2866 | mutex_unlock(&ext4_li_info->li_list_mtx); | |
2867 | } | |
2868 | ||
2869 | static int ext4_run_lazyinit_thread(void) | |
2870 | { | |
8f1f7453 ES |
2871 | ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, |
2872 | ext4_li_info, "ext4lazyinit"); | |
2873 | if (IS_ERR(ext4_lazyinit_task)) { | |
2874 | int err = PTR_ERR(ext4_lazyinit_task); | |
bfff6873 | 2875 | ext4_clear_request_list(); |
bfff6873 LC |
2876 | kfree(ext4_li_info); |
2877 | ext4_li_info = NULL; | |
2878 | printk(KERN_CRIT "EXT4: error %d creating inode table " | |
2879 | "initialization thread\n", | |
2880 | err); | |
2881 | return err; | |
2882 | } | |
2883 | ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; | |
bfff6873 LC |
2884 | return 0; |
2885 | } | |
2886 | ||
2887 | /* | |
2888 | * Check whether it make sense to run itable init. thread or not. | |
2889 | * If there is at least one uninitialized inode table, return | |
2890 | * corresponding group number, else the loop goes through all | |
2891 | * groups and return total number of groups. | |
2892 | */ | |
2893 | static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) | |
2894 | { | |
2895 | ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; | |
2896 | struct ext4_group_desc *gdp = NULL; | |
2897 | ||
2898 | for (group = 0; group < ngroups; group++) { | |
2899 | gdp = ext4_get_group_desc(sb, group, NULL); | |
2900 | if (!gdp) | |
2901 | continue; | |
2902 | ||
2903 | if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) | |
2904 | break; | |
2905 | } | |
2906 | ||
2907 | return group; | |
2908 | } | |
2909 | ||
2910 | static int ext4_li_info_new(void) | |
2911 | { | |
2912 | struct ext4_lazy_init *eli = NULL; | |
2913 | ||
2914 | eli = kzalloc(sizeof(*eli), GFP_KERNEL); | |
2915 | if (!eli) | |
2916 | return -ENOMEM; | |
2917 | ||
bfff6873 LC |
2918 | INIT_LIST_HEAD(&eli->li_request_list); |
2919 | mutex_init(&eli->li_list_mtx); | |
2920 | ||
bfff6873 LC |
2921 | eli->li_state |= EXT4_LAZYINIT_QUIT; |
2922 | ||
2923 | ext4_li_info = eli; | |
2924 | ||
2925 | return 0; | |
2926 | } | |
2927 | ||
2928 | static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, | |
2929 | ext4_group_t start) | |
2930 | { | |
2931 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
2932 | struct ext4_li_request *elr; | |
2933 | unsigned long rnd; | |
2934 | ||
2935 | elr = kzalloc(sizeof(*elr), GFP_KERNEL); | |
2936 | if (!elr) | |
2937 | return NULL; | |
2938 | ||
2939 | elr->lr_super = sb; | |
2940 | elr->lr_sbi = sbi; | |
2941 | elr->lr_next_group = start; | |
2942 | ||
2943 | /* | |
2944 | * Randomize first schedule time of the request to | |
2945 | * spread the inode table initialization requests | |
2946 | * better. | |
2947 | */ | |
2948 | get_random_bytes(&rnd, sizeof(rnd)); | |
2949 | elr->lr_next_sched = jiffies + (unsigned long)rnd % | |
2950 | (EXT4_DEF_LI_MAX_START_DELAY * HZ); | |
2951 | ||
2952 | return elr; | |
2953 | } | |
2954 | ||
2955 | static int ext4_register_li_request(struct super_block *sb, | |
2956 | ext4_group_t first_not_zeroed) | |
2957 | { | |
2958 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
2959 | struct ext4_li_request *elr; | |
2960 | ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; | |
6c5a6cb9 | 2961 | int ret = 0; |
bfff6873 | 2962 | |
51ce6511 LC |
2963 | if (sbi->s_li_request != NULL) { |
2964 | /* | |
2965 | * Reset timeout so it can be computed again, because | |
2966 | * s_li_wait_mult might have changed. | |
2967 | */ | |
2968 | sbi->s_li_request->lr_timeout = 0; | |
beed5ecb | 2969 | return 0; |
51ce6511 | 2970 | } |
bfff6873 LC |
2971 | |
2972 | if (first_not_zeroed == ngroups || | |
2973 | (sb->s_flags & MS_RDONLY) || | |
55ff3840 | 2974 | !test_opt(sb, INIT_INODE_TABLE)) |
beed5ecb | 2975 | return 0; |
bfff6873 LC |
2976 | |
2977 | elr = ext4_li_request_new(sb, first_not_zeroed); | |
beed5ecb NK |
2978 | if (!elr) |
2979 | return -ENOMEM; | |
bfff6873 LC |
2980 | |
2981 | mutex_lock(&ext4_li_mtx); | |
2982 | ||
2983 | if (NULL == ext4_li_info) { | |
2984 | ret = ext4_li_info_new(); | |
2985 | if (ret) | |
2986 | goto out; | |
2987 | } | |
2988 | ||
2989 | mutex_lock(&ext4_li_info->li_list_mtx); | |
2990 | list_add(&elr->lr_request, &ext4_li_info->li_request_list); | |
2991 | mutex_unlock(&ext4_li_info->li_list_mtx); | |
2992 | ||
2993 | sbi->s_li_request = elr; | |
46e4690b TM |
2994 | /* |
2995 | * set elr to NULL here since it has been inserted to | |
2996 | * the request_list and the removal and free of it is | |
2997 | * handled by ext4_clear_request_list from now on. | |
2998 | */ | |
2999 | elr = NULL; | |
bfff6873 LC |
3000 | |
3001 | if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { | |
3002 | ret = ext4_run_lazyinit_thread(); | |
3003 | if (ret) | |
3004 | goto out; | |
3005 | } | |
bfff6873 | 3006 | out: |
beed5ecb NK |
3007 | mutex_unlock(&ext4_li_mtx); |
3008 | if (ret) | |
bfff6873 | 3009 | kfree(elr); |
bfff6873 LC |
3010 | return ret; |
3011 | } | |
3012 | ||
3013 | /* | |
3014 | * We do not need to lock anything since this is called on | |
3015 | * module unload. | |
3016 | */ | |
3017 | static void ext4_destroy_lazyinit_thread(void) | |
3018 | { | |
3019 | /* | |
3020 | * If thread exited earlier | |
3021 | * there's nothing to be done. | |
3022 | */ | |
8f1f7453 | 3023 | if (!ext4_li_info || !ext4_lazyinit_task) |
bfff6873 LC |
3024 | return; |
3025 | ||
8f1f7453 | 3026 | kthread_stop(ext4_lazyinit_task); |
bfff6873 LC |
3027 | } |
3028 | ||
2b2d6d01 | 3029 | static int ext4_fill_super(struct super_block *sb, void *data, int silent) |
7477827f AK |
3030 | __releases(kernel_lock) |
3031 | __acquires(kernel_lock) | |
ac27a0ec | 3032 | { |
d4c402d9 | 3033 | char *orig_data = kstrdup(data, GFP_KERNEL); |
2b2d6d01 | 3034 | struct buffer_head *bh; |
617ba13b MC |
3035 | struct ext4_super_block *es = NULL; |
3036 | struct ext4_sb_info *sbi; | |
3037 | ext4_fsblk_t block; | |
3038 | ext4_fsblk_t sb_block = get_sb_block(&data); | |
70bbb3e0 | 3039 | ext4_fsblk_t logical_sb_block; |
ac27a0ec | 3040 | unsigned long offset = 0; |
ac27a0ec DK |
3041 | unsigned long journal_devnum = 0; |
3042 | unsigned long def_mount_opts; | |
3043 | struct inode *root; | |
9f6200bb | 3044 | char *cp; |
0390131b | 3045 | const char *descr; |
dcc7dae3 | 3046 | int ret = -ENOMEM; |
ac27a0ec | 3047 | int blocksize; |
4ec11028 TT |
3048 | unsigned int db_count; |
3049 | unsigned int i; | |
f287a1a5 | 3050 | int needs_recovery, has_huge_files; |
bd81d8ee | 3051 | __u64 blocks_count; |
833f4077 | 3052 | int err; |
b3881f74 | 3053 | unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; |
bfff6873 | 3054 | ext4_group_t first_not_zeroed; |
ac27a0ec DK |
3055 | |
3056 | sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); | |
3057 | if (!sbi) | |
dcc7dae3 | 3058 | goto out_free_orig; |
705895b6 PE |
3059 | |
3060 | sbi->s_blockgroup_lock = | |
3061 | kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); | |
3062 | if (!sbi->s_blockgroup_lock) { | |
3063 | kfree(sbi); | |
dcc7dae3 | 3064 | goto out_free_orig; |
705895b6 | 3065 | } |
ac27a0ec DK |
3066 | sb->s_fs_info = sbi; |
3067 | sbi->s_mount_opt = 0; | |
617ba13b MC |
3068 | sbi->s_resuid = EXT4_DEF_RESUID; |
3069 | sbi->s_resgid = EXT4_DEF_RESGID; | |
240799cd | 3070 | sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; |
d9c9bef1 | 3071 | sbi->s_sb_block = sb_block; |
f613dfcb TT |
3072 | if (sb->s_bdev->bd_part) |
3073 | sbi->s_sectors_written_start = | |
3074 | part_stat_read(sb->s_bdev->bd_part, sectors[1]); | |
ac27a0ec | 3075 | |
9f6200bb TT |
3076 | /* Cleanup superblock name */ |
3077 | for (cp = sb->s_id; (cp = strchr(cp, '/'));) | |
3078 | *cp = '!'; | |
3079 | ||
dcc7dae3 | 3080 | ret = -EINVAL; |
617ba13b | 3081 | blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); |
ac27a0ec | 3082 | if (!blocksize) { |
b31e1552 | 3083 | ext4_msg(sb, KERN_ERR, "unable to set blocksize"); |
ac27a0ec DK |
3084 | goto out_fail; |
3085 | } | |
3086 | ||
3087 | /* | |
617ba13b | 3088 | * The ext4 superblock will not be buffer aligned for other than 1kB |
ac27a0ec DK |
3089 | * block sizes. We need to calculate the offset from buffer start. |
3090 | */ | |
617ba13b | 3091 | if (blocksize != EXT4_MIN_BLOCK_SIZE) { |
70bbb3e0 AM |
3092 | logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; |
3093 | offset = do_div(logical_sb_block, blocksize); | |
ac27a0ec | 3094 | } else { |
70bbb3e0 | 3095 | logical_sb_block = sb_block; |
ac27a0ec DK |
3096 | } |
3097 | ||
70bbb3e0 | 3098 | if (!(bh = sb_bread(sb, logical_sb_block))) { |
b31e1552 | 3099 | ext4_msg(sb, KERN_ERR, "unable to read superblock"); |
ac27a0ec DK |
3100 | goto out_fail; |
3101 | } | |
3102 | /* | |
3103 | * Note: s_es must be initialized as soon as possible because | |
617ba13b | 3104 | * some ext4 macro-instructions depend on its value |
ac27a0ec | 3105 | */ |
617ba13b | 3106 | es = (struct ext4_super_block *) (((char *)bh->b_data) + offset); |
ac27a0ec DK |
3107 | sbi->s_es = es; |
3108 | sb->s_magic = le16_to_cpu(es->s_magic); | |
617ba13b MC |
3109 | if (sb->s_magic != EXT4_SUPER_MAGIC) |
3110 | goto cantfind_ext4; | |
afc32f7e | 3111 | sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); |
ac27a0ec DK |
3112 | |
3113 | /* Set defaults before we parse the mount options */ | |
3114 | def_mount_opts = le32_to_cpu(es->s_default_mount_opts); | |
fd8c37ec | 3115 | set_opt(sb, INIT_INODE_TABLE); |
617ba13b | 3116 | if (def_mount_opts & EXT4_DEFM_DEBUG) |
fd8c37ec | 3117 | set_opt(sb, DEBUG); |
437ca0fd DM |
3118 | if (def_mount_opts & EXT4_DEFM_BSDGROUPS) { |
3119 | ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups", | |
3120 | "2.6.38"); | |
fd8c37ec | 3121 | set_opt(sb, GRPID); |
437ca0fd | 3122 | } |
617ba13b | 3123 | if (def_mount_opts & EXT4_DEFM_UID16) |
fd8c37ec | 3124 | set_opt(sb, NO_UID32); |
ea663336 | 3125 | /* xattr user namespace & acls are now defaulted on */ |
03010a33 | 3126 | #ifdef CONFIG_EXT4_FS_XATTR |
ea663336 | 3127 | set_opt(sb, XATTR_USER); |
2e7842b8 | 3128 | #endif |
03010a33 | 3129 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
ea663336 | 3130 | set_opt(sb, POSIX_ACL); |
2e7842b8 | 3131 | #endif |
6fd7a467 | 3132 | set_opt(sb, MBLK_IO_SUBMIT); |
617ba13b | 3133 | if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) |
fd8c37ec | 3134 | set_opt(sb, JOURNAL_DATA); |
617ba13b | 3135 | else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) |
fd8c37ec | 3136 | set_opt(sb, ORDERED_DATA); |
617ba13b | 3137 | else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) |
fd8c37ec | 3138 | set_opt(sb, WRITEBACK_DATA); |
617ba13b MC |
3139 | |
3140 | if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) | |
fd8c37ec | 3141 | set_opt(sb, ERRORS_PANIC); |
bb4f397a | 3142 | else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) |
fd8c37ec | 3143 | set_opt(sb, ERRORS_CONT); |
bb4f397a | 3144 | else |
fd8c37ec | 3145 | set_opt(sb, ERRORS_RO); |
8b67f04a | 3146 | if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY) |
fd8c37ec | 3147 | set_opt(sb, BLOCK_VALIDITY); |
8b67f04a | 3148 | if (def_mount_opts & EXT4_DEFM_DISCARD) |
fd8c37ec | 3149 | set_opt(sb, DISCARD); |
ac27a0ec DK |
3150 | |
3151 | sbi->s_resuid = le16_to_cpu(es->s_def_resuid); | |
3152 | sbi->s_resgid = le16_to_cpu(es->s_def_resgid); | |
30773840 TT |
3153 | sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; |
3154 | sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; | |
3155 | sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; | |
ac27a0ec | 3156 | |
8b67f04a | 3157 | if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) |
fd8c37ec | 3158 | set_opt(sb, BARRIER); |
ac27a0ec | 3159 | |
dd919b98 AK |
3160 | /* |
3161 | * enable delayed allocation by default | |
3162 | * Use -o nodelalloc to turn it off | |
3163 | */ | |
8b67f04a TT |
3164 | if (!IS_EXT3_SB(sb) && |
3165 | ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) | |
fd8c37ec | 3166 | set_opt(sb, DELALLOC); |
dd919b98 | 3167 | |
51ce6511 LC |
3168 | /* |
3169 | * set default s_li_wait_mult for lazyinit, for the case there is | |
3170 | * no mount option specified. | |
3171 | */ | |
3172 | sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; | |
3173 | ||
8b67f04a TT |
3174 | if (!parse_options((char *) sbi->s_es->s_mount_opts, sb, |
3175 | &journal_devnum, &journal_ioprio, NULL, 0)) { | |
3176 | ext4_msg(sb, KERN_WARNING, | |
3177 | "failed to parse options in superblock: %s", | |
3178 | sbi->s_es->s_mount_opts); | |
3179 | } | |
b3881f74 TT |
3180 | if (!parse_options((char *) data, sb, &journal_devnum, |
3181 | &journal_ioprio, NULL, 0)) | |
ac27a0ec DK |
3182 | goto failed_mount; |
3183 | ||
3184 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
482a7425 | 3185 | (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); |
ac27a0ec | 3186 | |
617ba13b MC |
3187 | if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && |
3188 | (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) || | |
3189 | EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) || | |
3190 | EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U))) | |
b31e1552 ES |
3191 | ext4_msg(sb, KERN_WARNING, |
3192 | "feature flags set on rev 0 fs, " | |
3193 | "running e2fsck is recommended"); | |
469108ff | 3194 | |
2035e776 TT |
3195 | if (IS_EXT2_SB(sb)) { |
3196 | if (ext2_feature_set_ok(sb)) | |
3197 | ext4_msg(sb, KERN_INFO, "mounting ext2 file system " | |
3198 | "using the ext4 subsystem"); | |
3199 | else { | |
3200 | ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " | |
3201 | "to feature incompatibilities"); | |
3202 | goto failed_mount; | |
3203 | } | |
3204 | } | |
3205 | ||
3206 | if (IS_EXT3_SB(sb)) { | |
3207 | if (ext3_feature_set_ok(sb)) | |
3208 | ext4_msg(sb, KERN_INFO, "mounting ext3 file system " | |
3209 | "using the ext4 subsystem"); | |
3210 | else { | |
3211 | ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " | |
3212 | "to feature incompatibilities"); | |
3213 | goto failed_mount; | |
3214 | } | |
3215 | } | |
3216 | ||
ac27a0ec DK |
3217 | /* |
3218 | * Check feature flags regardless of the revision level, since we | |
3219 | * previously didn't change the revision level when setting the flags, | |
3220 | * so there is a chance incompat flags are set on a rev 0 filesystem. | |
3221 | */ | |
a13fb1a4 | 3222 | if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY))) |
ac27a0ec | 3223 | goto failed_mount; |
a13fb1a4 | 3224 | |
ac27a0ec DK |
3225 | blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); |
3226 | ||
617ba13b MC |
3227 | if (blocksize < EXT4_MIN_BLOCK_SIZE || |
3228 | blocksize > EXT4_MAX_BLOCK_SIZE) { | |
b31e1552 ES |
3229 | ext4_msg(sb, KERN_ERR, |
3230 | "Unsupported filesystem blocksize %d", blocksize); | |
ac27a0ec DK |
3231 | goto failed_mount; |
3232 | } | |
3233 | ||
ac27a0ec | 3234 | if (sb->s_blocksize != blocksize) { |
ce40733c AK |
3235 | /* Validate the filesystem blocksize */ |
3236 | if (!sb_set_blocksize(sb, blocksize)) { | |
b31e1552 | 3237 | ext4_msg(sb, KERN_ERR, "bad block size %d", |
ce40733c | 3238 | blocksize); |
ac27a0ec DK |
3239 | goto failed_mount; |
3240 | } | |
3241 | ||
2b2d6d01 | 3242 | brelse(bh); |
70bbb3e0 AM |
3243 | logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; |
3244 | offset = do_div(logical_sb_block, blocksize); | |
3245 | bh = sb_bread(sb, logical_sb_block); | |
ac27a0ec | 3246 | if (!bh) { |
b31e1552 ES |
3247 | ext4_msg(sb, KERN_ERR, |
3248 | "Can't read superblock on 2nd try"); | |
ac27a0ec DK |
3249 | goto failed_mount; |
3250 | } | |
617ba13b | 3251 | es = (struct ext4_super_block *)(((char *)bh->b_data) + offset); |
ac27a0ec | 3252 | sbi->s_es = es; |
617ba13b | 3253 | if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { |
b31e1552 ES |
3254 | ext4_msg(sb, KERN_ERR, |
3255 | "Magic mismatch, very weird!"); | |
ac27a0ec DK |
3256 | goto failed_mount; |
3257 | } | |
3258 | } | |
3259 | ||
a13fb1a4 ES |
3260 | has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb, |
3261 | EXT4_FEATURE_RO_COMPAT_HUGE_FILE); | |
f287a1a5 TT |
3262 | sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, |
3263 | has_huge_files); | |
3264 | sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); | |
ac27a0ec | 3265 | |
617ba13b MC |
3266 | if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { |
3267 | sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; | |
3268 | sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; | |
ac27a0ec DK |
3269 | } else { |
3270 | sbi->s_inode_size = le16_to_cpu(es->s_inode_size); | |
3271 | sbi->s_first_ino = le32_to_cpu(es->s_first_ino); | |
617ba13b | 3272 | if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || |
1330593e | 3273 | (!is_power_of_2(sbi->s_inode_size)) || |
ac27a0ec | 3274 | (sbi->s_inode_size > blocksize)) { |
b31e1552 ES |
3275 | ext4_msg(sb, KERN_ERR, |
3276 | "unsupported inode size: %d", | |
2b2d6d01 | 3277 | sbi->s_inode_size); |
ac27a0ec DK |
3278 | goto failed_mount; |
3279 | } | |
ef7f3835 KS |
3280 | if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) |
3281 | sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2); | |
ac27a0ec | 3282 | } |
0b8e58a1 | 3283 | |
0d1ee42f AR |
3284 | sbi->s_desc_size = le16_to_cpu(es->s_desc_size); |
3285 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) { | |
8fadc143 | 3286 | if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || |
0d1ee42f | 3287 | sbi->s_desc_size > EXT4_MAX_DESC_SIZE || |
d8ea6cf8 | 3288 | !is_power_of_2(sbi->s_desc_size)) { |
b31e1552 ES |
3289 | ext4_msg(sb, KERN_ERR, |
3290 | "unsupported descriptor size %lu", | |
0d1ee42f AR |
3291 | sbi->s_desc_size); |
3292 | goto failed_mount; | |
3293 | } | |
3294 | } else | |
3295 | sbi->s_desc_size = EXT4_MIN_DESC_SIZE; | |
0b8e58a1 | 3296 | |
ac27a0ec | 3297 | sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); |
ac27a0ec | 3298 | sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); |
b47b6f38 | 3299 | if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0) |
617ba13b | 3300 | goto cantfind_ext4; |
0b8e58a1 | 3301 | |
617ba13b | 3302 | sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); |
ac27a0ec | 3303 | if (sbi->s_inodes_per_block == 0) |
617ba13b | 3304 | goto cantfind_ext4; |
ac27a0ec DK |
3305 | sbi->s_itb_per_group = sbi->s_inodes_per_group / |
3306 | sbi->s_inodes_per_block; | |
0d1ee42f | 3307 | sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); |
ac27a0ec DK |
3308 | sbi->s_sbh = bh; |
3309 | sbi->s_mount_state = le16_to_cpu(es->s_state); | |
e57aa839 FW |
3310 | sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); |
3311 | sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); | |
0b8e58a1 | 3312 | |
2b2d6d01 | 3313 | for (i = 0; i < 4; i++) |
ac27a0ec DK |
3314 | sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); |
3315 | sbi->s_def_hash_version = es->s_def_hash_version; | |
f99b2589 TT |
3316 | i = le32_to_cpu(es->s_flags); |
3317 | if (i & EXT2_FLAGS_UNSIGNED_HASH) | |
3318 | sbi->s_hash_unsigned = 3; | |
3319 | else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { | |
3320 | #ifdef __CHAR_UNSIGNED__ | |
3321 | es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); | |
3322 | sbi->s_hash_unsigned = 3; | |
3323 | #else | |
3324 | es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); | |
3325 | #endif | |
3326 | sb->s_dirt = 1; | |
3327 | } | |
ac27a0ec DK |
3328 | |
3329 | if (sbi->s_blocks_per_group > blocksize * 8) { | |
b31e1552 ES |
3330 | ext4_msg(sb, KERN_ERR, |
3331 | "#blocks per group too big: %lu", | |
2b2d6d01 | 3332 | sbi->s_blocks_per_group); |
ac27a0ec DK |
3333 | goto failed_mount; |
3334 | } | |
ac27a0ec | 3335 | if (sbi->s_inodes_per_group > blocksize * 8) { |
b31e1552 ES |
3336 | ext4_msg(sb, KERN_ERR, |
3337 | "#inodes per group too big: %lu", | |
2b2d6d01 | 3338 | sbi->s_inodes_per_group); |
ac27a0ec DK |
3339 | goto failed_mount; |
3340 | } | |
3341 | ||
bf43d84b ES |
3342 | /* |
3343 | * Test whether we have more sectors than will fit in sector_t, | |
3344 | * and whether the max offset is addressable by the page cache. | |
3345 | */ | |
5a9ae68a | 3346 | err = generic_check_addressable(sb->s_blocksize_bits, |
30ca22c7 | 3347 | ext4_blocks_count(es)); |
5a9ae68a | 3348 | if (err) { |
b31e1552 | 3349 | ext4_msg(sb, KERN_ERR, "filesystem" |
bf43d84b | 3350 | " too large to mount safely on this system"); |
ac27a0ec | 3351 | if (sizeof(sector_t) < 8) |
90c699a9 | 3352 | ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled"); |
5a9ae68a | 3353 | ret = err; |
ac27a0ec DK |
3354 | goto failed_mount; |
3355 | } | |
3356 | ||
617ba13b MC |
3357 | if (EXT4_BLOCKS_PER_GROUP(sb) == 0) |
3358 | goto cantfind_ext4; | |
e7c95593 | 3359 | |
0f2ddca6 FTN |
3360 | /* check blocks count against device size */ |
3361 | blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; | |
3362 | if (blocks_count && ext4_blocks_count(es) > blocks_count) { | |
b31e1552 ES |
3363 | ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " |
3364 | "exceeds size of device (%llu blocks)", | |
0f2ddca6 FTN |
3365 | ext4_blocks_count(es), blocks_count); |
3366 | goto failed_mount; | |
3367 | } | |
3368 | ||
0b8e58a1 AD |
3369 | /* |
3370 | * It makes no sense for the first data block to be beyond the end | |
3371 | * of the filesystem. | |
3372 | */ | |
3373 | if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { | |
b31e1552 ES |
3374 | ext4_msg(sb, KERN_WARNING, "bad geometry: first data" |
3375 | "block %u is beyond end of filesystem (%llu)", | |
3376 | le32_to_cpu(es->s_first_data_block), | |
3377 | ext4_blocks_count(es)); | |
e7c95593 ES |
3378 | goto failed_mount; |
3379 | } | |
bd81d8ee LV |
3380 | blocks_count = (ext4_blocks_count(es) - |
3381 | le32_to_cpu(es->s_first_data_block) + | |
3382 | EXT4_BLOCKS_PER_GROUP(sb) - 1); | |
3383 | do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); | |
4ec11028 | 3384 | if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { |
b31e1552 | 3385 | ext4_msg(sb, KERN_WARNING, "groups count too large: %u " |
4ec11028 | 3386 | "(block count %llu, first data block %u, " |
b31e1552 | 3387 | "blocks per group %lu)", sbi->s_groups_count, |
4ec11028 TT |
3388 | ext4_blocks_count(es), |
3389 | le32_to_cpu(es->s_first_data_block), | |
3390 | EXT4_BLOCKS_PER_GROUP(sb)); | |
3391 | goto failed_mount; | |
3392 | } | |
bd81d8ee | 3393 | sbi->s_groups_count = blocks_count; |
fb0a387d ES |
3394 | sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, |
3395 | (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); | |
617ba13b MC |
3396 | db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / |
3397 | EXT4_DESC_PER_BLOCK(sb); | |
2b2d6d01 | 3398 | sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *), |
ac27a0ec DK |
3399 | GFP_KERNEL); |
3400 | if (sbi->s_group_desc == NULL) { | |
b31e1552 | 3401 | ext4_msg(sb, KERN_ERR, "not enough memory"); |
ac27a0ec DK |
3402 | goto failed_mount; |
3403 | } | |
3404 | ||
3244fcb1 | 3405 | #ifdef CONFIG_PROC_FS |
9f6200bb TT |
3406 | if (ext4_proc_root) |
3407 | sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root); | |
3244fcb1 | 3408 | #endif |
240799cd | 3409 | |
705895b6 | 3410 | bgl_lock_init(sbi->s_blockgroup_lock); |
ac27a0ec DK |
3411 | |
3412 | for (i = 0; i < db_count; i++) { | |
70bbb3e0 | 3413 | block = descriptor_loc(sb, logical_sb_block, i); |
ac27a0ec DK |
3414 | sbi->s_group_desc[i] = sb_bread(sb, block); |
3415 | if (!sbi->s_group_desc[i]) { | |
b31e1552 ES |
3416 | ext4_msg(sb, KERN_ERR, |
3417 | "can't read group descriptor %d", i); | |
ac27a0ec DK |
3418 | db_count = i; |
3419 | goto failed_mount2; | |
3420 | } | |
3421 | } | |
bfff6873 | 3422 | if (!ext4_check_descriptors(sb, &first_not_zeroed)) { |
b31e1552 | 3423 | ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); |
ac27a0ec DK |
3424 | goto failed_mount2; |
3425 | } | |
772cb7c8 JS |
3426 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) |
3427 | if (!ext4_fill_flex_info(sb)) { | |
b31e1552 ES |
3428 | ext4_msg(sb, KERN_ERR, |
3429 | "unable to initialize " | |
3430 | "flex_bg meta info!"); | |
772cb7c8 JS |
3431 | goto failed_mount2; |
3432 | } | |
3433 | ||
ac27a0ec DK |
3434 | sbi->s_gdb_count = db_count; |
3435 | get_random_bytes(&sbi->s_next_generation, sizeof(u32)); | |
3436 | spin_lock_init(&sbi->s_next_gen_lock); | |
3437 | ||
04496411 TM |
3438 | init_timer(&sbi->s_err_report); |
3439 | sbi->s_err_report.function = print_daily_error_info; | |
3440 | sbi->s_err_report.data = (unsigned long) sb; | |
3441 | ||
ce7e010a TT |
3442 | err = percpu_counter_init(&sbi->s_freeblocks_counter, |
3443 | ext4_count_free_blocks(sb)); | |
3444 | if (!err) { | |
3445 | err = percpu_counter_init(&sbi->s_freeinodes_counter, | |
3446 | ext4_count_free_inodes(sb)); | |
3447 | } | |
3448 | if (!err) { | |
3449 | err = percpu_counter_init(&sbi->s_dirs_counter, | |
3450 | ext4_count_dirs(sb)); | |
3451 | } | |
3452 | if (!err) { | |
3453 | err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0); | |
3454 | } | |
3455 | if (err) { | |
3456 | ext4_msg(sb, KERN_ERR, "insufficient memory"); | |
3457 | goto failed_mount3; | |
3458 | } | |
3459 | ||
c9de560d | 3460 | sbi->s_stripe = ext4_get_stripe_size(sbi); |
55138e0b | 3461 | sbi->s_max_writeback_mb_bump = 128; |
c9de560d | 3462 | |
ac27a0ec DK |
3463 | /* |
3464 | * set up enough so that it can read an inode | |
3465 | */ | |
9ca92389 TT |
3466 | if (!test_opt(sb, NOLOAD) && |
3467 | EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) | |
3468 | sb->s_op = &ext4_sops; | |
3469 | else | |
3470 | sb->s_op = &ext4_nojournal_sops; | |
617ba13b MC |
3471 | sb->s_export_op = &ext4_export_ops; |
3472 | sb->s_xattr = ext4_xattr_handlers; | |
ac27a0ec | 3473 | #ifdef CONFIG_QUOTA |
617ba13b MC |
3474 | sb->s_qcop = &ext4_qctl_operations; |
3475 | sb->dq_op = &ext4_quota_operations; | |
ac27a0ec | 3476 | #endif |
f2fa2ffc AK |
3477 | memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); |
3478 | ||
ac27a0ec | 3479 | INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ |
3b9d4ed2 | 3480 | mutex_init(&sbi->s_orphan_lock); |
32ed5058 | 3481 | mutex_init(&sbi->s_resize_lock); |
ac27a0ec DK |
3482 | |
3483 | sb->s_root = NULL; | |
3484 | ||
3485 | needs_recovery = (es->s_last_orphan != 0 || | |
617ba13b MC |
3486 | EXT4_HAS_INCOMPAT_FEATURE(sb, |
3487 | EXT4_FEATURE_INCOMPAT_RECOVER)); | |
ac27a0ec DK |
3488 | |
3489 | /* | |
3490 | * The first inode we look at is the journal inode. Don't try | |
3491 | * root first: it may be modified in the journal! | |
3492 | */ | |
3493 | if (!test_opt(sb, NOLOAD) && | |
617ba13b MC |
3494 | EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { |
3495 | if (ext4_load_journal(sb, es, journal_devnum)) | |
ac27a0ec | 3496 | goto failed_mount3; |
0390131b FM |
3497 | } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) && |
3498 | EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { | |
b31e1552 ES |
3499 | ext4_msg(sb, KERN_ERR, "required journal recovery " |
3500 | "suppressed and not mounted read-only"); | |
744692dc | 3501 | goto failed_mount_wq; |
ac27a0ec | 3502 | } else { |
fd8c37ec | 3503 | clear_opt(sb, DATA_FLAGS); |
0390131b FM |
3504 | sbi->s_journal = NULL; |
3505 | needs_recovery = 0; | |
3506 | goto no_journal; | |
ac27a0ec DK |
3507 | } |
3508 | ||
eb40a09c JS |
3509 | if (ext4_blocks_count(es) > 0xffffffffULL && |
3510 | !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, | |
3511 | JBD2_FEATURE_INCOMPAT_64BIT)) { | |
b31e1552 | 3512 | ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); |
744692dc | 3513 | goto failed_mount_wq; |
eb40a09c JS |
3514 | } |
3515 | ||
d4da6c9c LT |
3516 | if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { |
3517 | jbd2_journal_set_features(sbi->s_journal, | |
3518 | JBD2_FEATURE_COMPAT_CHECKSUM, 0, | |
818d276c | 3519 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); |
d4da6c9c LT |
3520 | } else if (test_opt(sb, JOURNAL_CHECKSUM)) { |
3521 | jbd2_journal_set_features(sbi->s_journal, | |
3522 | JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0); | |
818d276c GS |
3523 | jbd2_journal_clear_features(sbi->s_journal, 0, 0, |
3524 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); | |
d4da6c9c LT |
3525 | } else { |
3526 | jbd2_journal_clear_features(sbi->s_journal, | |
3527 | JBD2_FEATURE_COMPAT_CHECKSUM, 0, | |
3528 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); | |
3529 | } | |
818d276c | 3530 | |
ac27a0ec DK |
3531 | /* We have now updated the journal if required, so we can |
3532 | * validate the data journaling mode. */ | |
3533 | switch (test_opt(sb, DATA_FLAGS)) { | |
3534 | case 0: | |
3535 | /* No mode set, assume a default based on the journal | |
63f57933 AM |
3536 | * capabilities: ORDERED_DATA if the journal can |
3537 | * cope, else JOURNAL_DATA | |
3538 | */ | |
dab291af MC |
3539 | if (jbd2_journal_check_available_features |
3540 | (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) | |
fd8c37ec | 3541 | set_opt(sb, ORDERED_DATA); |
ac27a0ec | 3542 | else |
fd8c37ec | 3543 | set_opt(sb, JOURNAL_DATA); |
ac27a0ec DK |
3544 | break; |
3545 | ||
617ba13b MC |
3546 | case EXT4_MOUNT_ORDERED_DATA: |
3547 | case EXT4_MOUNT_WRITEBACK_DATA: | |
dab291af MC |
3548 | if (!jbd2_journal_check_available_features |
3549 | (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { | |
b31e1552 ES |
3550 | ext4_msg(sb, KERN_ERR, "Journal does not support " |
3551 | "requested data journaling mode"); | |
744692dc | 3552 | goto failed_mount_wq; |
ac27a0ec DK |
3553 | } |
3554 | default: | |
3555 | break; | |
3556 | } | |
b3881f74 | 3557 | set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); |
ac27a0ec | 3558 | |
ce7e010a TT |
3559 | /* |
3560 | * The journal may have updated the bg summary counts, so we | |
3561 | * need to update the global counters. | |
3562 | */ | |
3563 | percpu_counter_set(&sbi->s_freeblocks_counter, | |
3564 | ext4_count_free_blocks(sb)); | |
3565 | percpu_counter_set(&sbi->s_freeinodes_counter, | |
3566 | ext4_count_free_inodes(sb)); | |
3567 | percpu_counter_set(&sbi->s_dirs_counter, | |
3568 | ext4_count_dirs(sb)); | |
3569 | percpu_counter_set(&sbi->s_dirtyblocks_counter, 0); | |
206f7ab4 | 3570 | |
ce7e010a | 3571 | no_journal: |
fd89d5f2 TH |
3572 | /* |
3573 | * The maximum number of concurrent works can be high and | |
3574 | * concurrency isn't really necessary. Limit it to 1. | |
3575 | */ | |
3576 | EXT4_SB(sb)->dio_unwritten_wq = | |
ae005cbe | 3577 | alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); |
4c0425ff MC |
3578 | if (!EXT4_SB(sb)->dio_unwritten_wq) { |
3579 | printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n"); | |
3580 | goto failed_mount_wq; | |
3581 | } | |
3582 | ||
ac27a0ec | 3583 | /* |
dab291af | 3584 | * The jbd2_journal_load will have done any necessary log recovery, |
ac27a0ec DK |
3585 | * so we can safely mount the rest of the filesystem now. |
3586 | */ | |
3587 | ||
1d1fe1ee DH |
3588 | root = ext4_iget(sb, EXT4_ROOT_INO); |
3589 | if (IS_ERR(root)) { | |
b31e1552 | 3590 | ext4_msg(sb, KERN_ERR, "get root inode failed"); |
1d1fe1ee | 3591 | ret = PTR_ERR(root); |
32a9bb57 | 3592 | root = NULL; |
ac27a0ec DK |
3593 | goto failed_mount4; |
3594 | } | |
3595 | if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { | |
b31e1552 | 3596 | ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); |
ac27a0ec DK |
3597 | goto failed_mount4; |
3598 | } | |
1d1fe1ee DH |
3599 | sb->s_root = d_alloc_root(root); |
3600 | if (!sb->s_root) { | |
b31e1552 | 3601 | ext4_msg(sb, KERN_ERR, "get root dentry failed"); |
1d1fe1ee DH |
3602 | ret = -ENOMEM; |
3603 | goto failed_mount4; | |
3604 | } | |
ac27a0ec | 3605 | |
2b2d6d01 | 3606 | ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY); |
ef7f3835 KS |
3607 | |
3608 | /* determine the minimum size of new large inodes, if present */ | |
3609 | if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { | |
3610 | sbi->s_want_extra_isize = sizeof(struct ext4_inode) - | |
3611 | EXT4_GOOD_OLD_INODE_SIZE; | |
3612 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, | |
3613 | EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) { | |
3614 | if (sbi->s_want_extra_isize < | |
3615 | le16_to_cpu(es->s_want_extra_isize)) | |
3616 | sbi->s_want_extra_isize = | |
3617 | le16_to_cpu(es->s_want_extra_isize); | |
3618 | if (sbi->s_want_extra_isize < | |
3619 | le16_to_cpu(es->s_min_extra_isize)) | |
3620 | sbi->s_want_extra_isize = | |
3621 | le16_to_cpu(es->s_min_extra_isize); | |
3622 | } | |
3623 | } | |
3624 | /* Check if enough inode space is available */ | |
3625 | if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize > | |
3626 | sbi->s_inode_size) { | |
3627 | sbi->s_want_extra_isize = sizeof(struct ext4_inode) - | |
3628 | EXT4_GOOD_OLD_INODE_SIZE; | |
b31e1552 ES |
3629 | ext4_msg(sb, KERN_INFO, "required extra inode space not" |
3630 | "available"); | |
ef7f3835 KS |
3631 | } |
3632 | ||
90576c0b TT |
3633 | if (test_opt(sb, DELALLOC) && |
3634 | (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) { | |
b31e1552 ES |
3635 | ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - " |
3636 | "requested data journaling mode"); | |
fd8c37ec | 3637 | clear_opt(sb, DELALLOC); |
90576c0b | 3638 | } |
744692dc JZ |
3639 | if (test_opt(sb, DIOREAD_NOLOCK)) { |
3640 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { | |
3641 | ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock " | |
3642 | "option - requested data journaling mode"); | |
fd8c37ec | 3643 | clear_opt(sb, DIOREAD_NOLOCK); |
744692dc JZ |
3644 | } |
3645 | if (sb->s_blocksize < PAGE_SIZE) { | |
3646 | ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock " | |
3647 | "option - block size is too small"); | |
fd8c37ec | 3648 | clear_opt(sb, DIOREAD_NOLOCK); |
744692dc JZ |
3649 | } |
3650 | } | |
c2774d84 | 3651 | |
6fd058f7 TT |
3652 | err = ext4_setup_system_zone(sb); |
3653 | if (err) { | |
b31e1552 | 3654 | ext4_msg(sb, KERN_ERR, "failed to initialize system " |
fbe845dd | 3655 | "zone (%d)", err); |
6fd058f7 TT |
3656 | goto failed_mount4; |
3657 | } | |
3658 | ||
c2774d84 AK |
3659 | ext4_ext_init(sb); |
3660 | err = ext4_mb_init(sb, needs_recovery); | |
3661 | if (err) { | |
421f91d2 | 3662 | ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", |
b31e1552 | 3663 | err); |
c2774d84 AK |
3664 | goto failed_mount4; |
3665 | } | |
3666 | ||
bfff6873 LC |
3667 | err = ext4_register_li_request(sb, first_not_zeroed); |
3668 | if (err) | |
3669 | goto failed_mount4; | |
3670 | ||
3197ebdb TT |
3671 | sbi->s_kobj.kset = ext4_kset; |
3672 | init_completion(&sbi->s_kobj_unregister); | |
3673 | err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL, | |
3674 | "%s", sb->s_id); | |
3675 | if (err) { | |
3676 | ext4_mb_release(sb); | |
3677 | ext4_ext_release(sb); | |
3678 | goto failed_mount4; | |
3679 | }; | |
3680 | ||
617ba13b MC |
3681 | EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; |
3682 | ext4_orphan_cleanup(sb, es); | |
3683 | EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; | |
0390131b | 3684 | if (needs_recovery) { |
b31e1552 | 3685 | ext4_msg(sb, KERN_INFO, "recovery complete"); |
0390131b FM |
3686 | ext4_mark_recovery_complete(sb, es); |
3687 | } | |
3688 | if (EXT4_SB(sb)->s_journal) { | |
3689 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) | |
3690 | descr = " journalled data mode"; | |
3691 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) | |
3692 | descr = " ordered data mode"; | |
3693 | else | |
3694 | descr = " writeback data mode"; | |
3695 | } else | |
3696 | descr = "out journal"; | |
3697 | ||
d4c402d9 | 3698 | ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " |
8b67f04a TT |
3699 | "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts, |
3700 | *sbi->s_es->s_mount_opts ? "; " : "", orig_data); | |
ac27a0ec | 3701 | |
66e61a9e TT |
3702 | if (es->s_error_count) |
3703 | mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ | |
ac27a0ec | 3704 | |
d4c402d9 | 3705 | kfree(orig_data); |
ac27a0ec DK |
3706 | return 0; |
3707 | ||
617ba13b | 3708 | cantfind_ext4: |
ac27a0ec | 3709 | if (!silent) |
b31e1552 | 3710 | ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); |
ac27a0ec DK |
3711 | goto failed_mount; |
3712 | ||
3713 | failed_mount4: | |
32a9bb57 MK |
3714 | iput(root); |
3715 | sb->s_root = NULL; | |
b31e1552 | 3716 | ext4_msg(sb, KERN_ERR, "mount failed"); |
4c0425ff MC |
3717 | destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq); |
3718 | failed_mount_wq: | |
6fd058f7 | 3719 | ext4_release_system_zone(sb); |
0390131b FM |
3720 | if (sbi->s_journal) { |
3721 | jbd2_journal_destroy(sbi->s_journal); | |
3722 | sbi->s_journal = NULL; | |
3723 | } | |
ac27a0ec | 3724 | failed_mount3: |
04496411 | 3725 | del_timer(&sbi->s_err_report); |
c5ca7c76 TT |
3726 | if (sbi->s_flex_groups) { |
3727 | if (is_vmalloc_addr(sbi->s_flex_groups)) | |
3728 | vfree(sbi->s_flex_groups); | |
3729 | else | |
3730 | kfree(sbi->s_flex_groups); | |
3731 | } | |
ce7e010a TT |
3732 | percpu_counter_destroy(&sbi->s_freeblocks_counter); |
3733 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
3734 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
3735 | percpu_counter_destroy(&sbi->s_dirtyblocks_counter); | |
ac27a0ec DK |
3736 | failed_mount2: |
3737 | for (i = 0; i < db_count; i++) | |
3738 | brelse(sbi->s_group_desc[i]); | |
3739 | kfree(sbi->s_group_desc); | |
3740 | failed_mount: | |
240799cd | 3741 | if (sbi->s_proc) { |
9f6200bb | 3742 | remove_proc_entry(sb->s_id, ext4_proc_root); |
240799cd | 3743 | } |
ac27a0ec DK |
3744 | #ifdef CONFIG_QUOTA |
3745 | for (i = 0; i < MAXQUOTAS; i++) | |
3746 | kfree(sbi->s_qf_names[i]); | |
3747 | #endif | |
617ba13b | 3748 | ext4_blkdev_remove(sbi); |
ac27a0ec DK |
3749 | brelse(bh); |
3750 | out_fail: | |
3751 | sb->s_fs_info = NULL; | |
f6830165 | 3752 | kfree(sbi->s_blockgroup_lock); |
ac27a0ec | 3753 | kfree(sbi); |
dcc7dae3 | 3754 | out_free_orig: |
d4c402d9 | 3755 | kfree(orig_data); |
1d1fe1ee | 3756 | return ret; |
ac27a0ec DK |
3757 | } |
3758 | ||
3759 | /* | |
3760 | * Setup any per-fs journal parameters now. We'll do this both on | |
3761 | * initial mount, once the journal has been initialised but before we've | |
3762 | * done any recovery; and again on any subsequent remount. | |
3763 | */ | |
617ba13b | 3764 | static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) |
ac27a0ec | 3765 | { |
617ba13b | 3766 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 3767 | |
30773840 TT |
3768 | journal->j_commit_interval = sbi->s_commit_interval; |
3769 | journal->j_min_batch_time = sbi->s_min_batch_time; | |
3770 | journal->j_max_batch_time = sbi->s_max_batch_time; | |
ac27a0ec | 3771 | |
a931da6a | 3772 | write_lock(&journal->j_state_lock); |
ac27a0ec | 3773 | if (test_opt(sb, BARRIER)) |
dab291af | 3774 | journal->j_flags |= JBD2_BARRIER; |
ac27a0ec | 3775 | else |
dab291af | 3776 | journal->j_flags &= ~JBD2_BARRIER; |
5bf5683a HK |
3777 | if (test_opt(sb, DATA_ERR_ABORT)) |
3778 | journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; | |
3779 | else | |
3780 | journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; | |
a931da6a | 3781 | write_unlock(&journal->j_state_lock); |
ac27a0ec DK |
3782 | } |
3783 | ||
617ba13b | 3784 | static journal_t *ext4_get_journal(struct super_block *sb, |
ac27a0ec DK |
3785 | unsigned int journal_inum) |
3786 | { | |
3787 | struct inode *journal_inode; | |
3788 | journal_t *journal; | |
3789 | ||
0390131b FM |
3790 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
3791 | ||
ac27a0ec DK |
3792 | /* First, test for the existence of a valid inode on disk. Bad |
3793 | * things happen if we iget() an unused inode, as the subsequent | |
3794 | * iput() will try to delete it. */ | |
3795 | ||
1d1fe1ee DH |
3796 | journal_inode = ext4_iget(sb, journal_inum); |
3797 | if (IS_ERR(journal_inode)) { | |
b31e1552 | 3798 | ext4_msg(sb, KERN_ERR, "no journal found"); |
ac27a0ec DK |
3799 | return NULL; |
3800 | } | |
3801 | if (!journal_inode->i_nlink) { | |
3802 | make_bad_inode(journal_inode); | |
3803 | iput(journal_inode); | |
b31e1552 | 3804 | ext4_msg(sb, KERN_ERR, "journal inode is deleted"); |
ac27a0ec DK |
3805 | return NULL; |
3806 | } | |
3807 | ||
e5f8eab8 | 3808 | jbd_debug(2, "Journal inode found at %p: %lld bytes\n", |
ac27a0ec | 3809 | journal_inode, journal_inode->i_size); |
1d1fe1ee | 3810 | if (!S_ISREG(journal_inode->i_mode)) { |
b31e1552 | 3811 | ext4_msg(sb, KERN_ERR, "invalid journal inode"); |
ac27a0ec DK |
3812 | iput(journal_inode); |
3813 | return NULL; | |
3814 | } | |
3815 | ||
dab291af | 3816 | journal = jbd2_journal_init_inode(journal_inode); |
ac27a0ec | 3817 | if (!journal) { |
b31e1552 | 3818 | ext4_msg(sb, KERN_ERR, "Could not load journal inode"); |
ac27a0ec DK |
3819 | iput(journal_inode); |
3820 | return NULL; | |
3821 | } | |
3822 | journal->j_private = sb; | |
617ba13b | 3823 | ext4_init_journal_params(sb, journal); |
ac27a0ec DK |
3824 | return journal; |
3825 | } | |
3826 | ||
617ba13b | 3827 | static journal_t *ext4_get_dev_journal(struct super_block *sb, |
ac27a0ec DK |
3828 | dev_t j_dev) |
3829 | { | |
2b2d6d01 | 3830 | struct buffer_head *bh; |
ac27a0ec | 3831 | journal_t *journal; |
617ba13b MC |
3832 | ext4_fsblk_t start; |
3833 | ext4_fsblk_t len; | |
ac27a0ec | 3834 | int hblock, blocksize; |
617ba13b | 3835 | ext4_fsblk_t sb_block; |
ac27a0ec | 3836 | unsigned long offset; |
2b2d6d01 | 3837 | struct ext4_super_block *es; |
ac27a0ec DK |
3838 | struct block_device *bdev; |
3839 | ||
0390131b FM |
3840 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
3841 | ||
b31e1552 | 3842 | bdev = ext4_blkdev_get(j_dev, sb); |
ac27a0ec DK |
3843 | if (bdev == NULL) |
3844 | return NULL; | |
3845 | ||
ac27a0ec | 3846 | blocksize = sb->s_blocksize; |
e1defc4f | 3847 | hblock = bdev_logical_block_size(bdev); |
ac27a0ec | 3848 | if (blocksize < hblock) { |
b31e1552 ES |
3849 | ext4_msg(sb, KERN_ERR, |
3850 | "blocksize too small for journal device"); | |
ac27a0ec DK |
3851 | goto out_bdev; |
3852 | } | |
3853 | ||
617ba13b MC |
3854 | sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; |
3855 | offset = EXT4_MIN_BLOCK_SIZE % blocksize; | |
ac27a0ec DK |
3856 | set_blocksize(bdev, blocksize); |
3857 | if (!(bh = __bread(bdev, sb_block, blocksize))) { | |
b31e1552 ES |
3858 | ext4_msg(sb, KERN_ERR, "couldn't read superblock of " |
3859 | "external journal"); | |
ac27a0ec DK |
3860 | goto out_bdev; |
3861 | } | |
3862 | ||
617ba13b MC |
3863 | es = (struct ext4_super_block *) (((char *)bh->b_data) + offset); |
3864 | if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || | |
ac27a0ec | 3865 | !(le32_to_cpu(es->s_feature_incompat) & |
617ba13b | 3866 | EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { |
b31e1552 ES |
3867 | ext4_msg(sb, KERN_ERR, "external journal has " |
3868 | "bad superblock"); | |
ac27a0ec DK |
3869 | brelse(bh); |
3870 | goto out_bdev; | |
3871 | } | |
3872 | ||
617ba13b | 3873 | if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { |
b31e1552 | 3874 | ext4_msg(sb, KERN_ERR, "journal UUID does not match"); |
ac27a0ec DK |
3875 | brelse(bh); |
3876 | goto out_bdev; | |
3877 | } | |
3878 | ||
bd81d8ee | 3879 | len = ext4_blocks_count(es); |
ac27a0ec DK |
3880 | start = sb_block + 1; |
3881 | brelse(bh); /* we're done with the superblock */ | |
3882 | ||
dab291af | 3883 | journal = jbd2_journal_init_dev(bdev, sb->s_bdev, |
ac27a0ec DK |
3884 | start, len, blocksize); |
3885 | if (!journal) { | |
b31e1552 | 3886 | ext4_msg(sb, KERN_ERR, "failed to create device journal"); |
ac27a0ec DK |
3887 | goto out_bdev; |
3888 | } | |
3889 | journal->j_private = sb; | |
3890 | ll_rw_block(READ, 1, &journal->j_sb_buffer); | |
3891 | wait_on_buffer(journal->j_sb_buffer); | |
3892 | if (!buffer_uptodate(journal->j_sb_buffer)) { | |
b31e1552 | 3893 | ext4_msg(sb, KERN_ERR, "I/O error on journal device"); |
ac27a0ec DK |
3894 | goto out_journal; |
3895 | } | |
3896 | if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { | |
b31e1552 ES |
3897 | ext4_msg(sb, KERN_ERR, "External journal has more than one " |
3898 | "user (unsupported) - %d", | |
ac27a0ec DK |
3899 | be32_to_cpu(journal->j_superblock->s_nr_users)); |
3900 | goto out_journal; | |
3901 | } | |
617ba13b MC |
3902 | EXT4_SB(sb)->journal_bdev = bdev; |
3903 | ext4_init_journal_params(sb, journal); | |
ac27a0ec | 3904 | return journal; |
0b8e58a1 | 3905 | |
ac27a0ec | 3906 | out_journal: |
dab291af | 3907 | jbd2_journal_destroy(journal); |
ac27a0ec | 3908 | out_bdev: |
617ba13b | 3909 | ext4_blkdev_put(bdev); |
ac27a0ec DK |
3910 | return NULL; |
3911 | } | |
3912 | ||
617ba13b MC |
3913 | static int ext4_load_journal(struct super_block *sb, |
3914 | struct ext4_super_block *es, | |
ac27a0ec DK |
3915 | unsigned long journal_devnum) |
3916 | { | |
3917 | journal_t *journal; | |
3918 | unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); | |
3919 | dev_t journal_dev; | |
3920 | int err = 0; | |
3921 | int really_read_only; | |
3922 | ||
0390131b FM |
3923 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
3924 | ||
ac27a0ec DK |
3925 | if (journal_devnum && |
3926 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { | |
b31e1552 ES |
3927 | ext4_msg(sb, KERN_INFO, "external journal device major/minor " |
3928 | "numbers have changed"); | |
ac27a0ec DK |
3929 | journal_dev = new_decode_dev(journal_devnum); |
3930 | } else | |
3931 | journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); | |
3932 | ||
3933 | really_read_only = bdev_read_only(sb->s_bdev); | |
3934 | ||
3935 | /* | |
3936 | * Are we loading a blank journal or performing recovery after a | |
3937 | * crash? For recovery, we need to check in advance whether we | |
3938 | * can get read-write access to the device. | |
3939 | */ | |
617ba13b | 3940 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { |
ac27a0ec | 3941 | if (sb->s_flags & MS_RDONLY) { |
b31e1552 ES |
3942 | ext4_msg(sb, KERN_INFO, "INFO: recovery " |
3943 | "required on readonly filesystem"); | |
ac27a0ec | 3944 | if (really_read_only) { |
b31e1552 ES |
3945 | ext4_msg(sb, KERN_ERR, "write access " |
3946 | "unavailable, cannot proceed"); | |
ac27a0ec DK |
3947 | return -EROFS; |
3948 | } | |
b31e1552 ES |
3949 | ext4_msg(sb, KERN_INFO, "write access will " |
3950 | "be enabled during recovery"); | |
ac27a0ec DK |
3951 | } |
3952 | } | |
3953 | ||
3954 | if (journal_inum && journal_dev) { | |
b31e1552 ES |
3955 | ext4_msg(sb, KERN_ERR, "filesystem has both journal " |
3956 | "and inode journals!"); | |
ac27a0ec DK |
3957 | return -EINVAL; |
3958 | } | |
3959 | ||
3960 | if (journal_inum) { | |
617ba13b | 3961 | if (!(journal = ext4_get_journal(sb, journal_inum))) |
ac27a0ec DK |
3962 | return -EINVAL; |
3963 | } else { | |
617ba13b | 3964 | if (!(journal = ext4_get_dev_journal(sb, journal_dev))) |
ac27a0ec DK |
3965 | return -EINVAL; |
3966 | } | |
3967 | ||
90576c0b | 3968 | if (!(journal->j_flags & JBD2_BARRIER)) |
b31e1552 | 3969 | ext4_msg(sb, KERN_INFO, "barriers disabled"); |
4776004f | 3970 | |
ac27a0ec | 3971 | if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) { |
dab291af | 3972 | err = jbd2_journal_update_format(journal); |
ac27a0ec | 3973 | if (err) { |
b31e1552 | 3974 | ext4_msg(sb, KERN_ERR, "error updating journal"); |
dab291af | 3975 | jbd2_journal_destroy(journal); |
ac27a0ec DK |
3976 | return err; |
3977 | } | |
3978 | } | |
3979 | ||
617ba13b | 3980 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) |
dab291af | 3981 | err = jbd2_journal_wipe(journal, !really_read_only); |
1c13d5c0 TT |
3982 | if (!err) { |
3983 | char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); | |
3984 | if (save) | |
3985 | memcpy(save, ((char *) es) + | |
3986 | EXT4_S_ERR_START, EXT4_S_ERR_LEN); | |
dab291af | 3987 | err = jbd2_journal_load(journal); |
1c13d5c0 TT |
3988 | if (save) |
3989 | memcpy(((char *) es) + EXT4_S_ERR_START, | |
3990 | save, EXT4_S_ERR_LEN); | |
3991 | kfree(save); | |
3992 | } | |
ac27a0ec DK |
3993 | |
3994 | if (err) { | |
b31e1552 | 3995 | ext4_msg(sb, KERN_ERR, "error loading journal"); |
dab291af | 3996 | jbd2_journal_destroy(journal); |
ac27a0ec DK |
3997 | return err; |
3998 | } | |
3999 | ||
617ba13b MC |
4000 | EXT4_SB(sb)->s_journal = journal; |
4001 | ext4_clear_journal_err(sb, es); | |
ac27a0ec | 4002 | |
c41303ce | 4003 | if (!really_read_only && journal_devnum && |
ac27a0ec DK |
4004 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { |
4005 | es->s_journal_dev = cpu_to_le32(journal_devnum); | |
ac27a0ec DK |
4006 | |
4007 | /* Make sure we flush the recovery flag to disk. */ | |
e2d67052 | 4008 | ext4_commit_super(sb, 1); |
ac27a0ec DK |
4009 | } |
4010 | ||
4011 | return 0; | |
4012 | } | |
4013 | ||
e2d67052 | 4014 | static int ext4_commit_super(struct super_block *sb, int sync) |
ac27a0ec | 4015 | { |
e2d67052 | 4016 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
617ba13b | 4017 | struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; |
c4be0c1d | 4018 | int error = 0; |
ac27a0ec DK |
4019 | |
4020 | if (!sbh) | |
c4be0c1d | 4021 | return error; |
914258bf TT |
4022 | if (buffer_write_io_error(sbh)) { |
4023 | /* | |
4024 | * Oh, dear. A previous attempt to write the | |
4025 | * superblock failed. This could happen because the | |
4026 | * USB device was yanked out. Or it could happen to | |
4027 | * be a transient write error and maybe the block will | |
4028 | * be remapped. Nothing we can do but to retry the | |
4029 | * write and hope for the best. | |
4030 | */ | |
b31e1552 ES |
4031 | ext4_msg(sb, KERN_ERR, "previous I/O error to " |
4032 | "superblock detected"); | |
914258bf TT |
4033 | clear_buffer_write_io_error(sbh); |
4034 | set_buffer_uptodate(sbh); | |
4035 | } | |
71290b36 TT |
4036 | /* |
4037 | * If the file system is mounted read-only, don't update the | |
4038 | * superblock write time. This avoids updating the superblock | |
4039 | * write time when we are mounting the root file system | |
4040 | * read/only but we need to replay the journal; at that point, | |
4041 | * for people who are east of GMT and who make their clock | |
4042 | * tick in localtime for Windows bug-for-bug compatibility, | |
4043 | * the clock is set in the future, and this will cause e2fsck | |
4044 | * to complain and force a full file system check. | |
4045 | */ | |
4046 | if (!(sb->s_flags & MS_RDONLY)) | |
4047 | es->s_wtime = cpu_to_le32(get_seconds()); | |
f613dfcb TT |
4048 | if (sb->s_bdev->bd_part) |
4049 | es->s_kbytes_written = | |
4050 | cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + | |
afc32f7e TT |
4051 | ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - |
4052 | EXT4_SB(sb)->s_sectors_written_start) >> 1)); | |
f613dfcb TT |
4053 | else |
4054 | es->s_kbytes_written = | |
4055 | cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); | |
ce7e010a TT |
4056 | ext4_free_blocks_count_set(es, percpu_counter_sum_positive( |
4057 | &EXT4_SB(sb)->s_freeblocks_counter)); | |
4058 | es->s_free_inodes_count = | |
4059 | cpu_to_le32(percpu_counter_sum_positive( | |
4060 | &EXT4_SB(sb)->s_freeinodes_counter)); | |
7234ab2a | 4061 | sb->s_dirt = 0; |
ac27a0ec DK |
4062 | BUFFER_TRACE(sbh, "marking dirty"); |
4063 | mark_buffer_dirty(sbh); | |
914258bf | 4064 | if (sync) { |
c4be0c1d TS |
4065 | error = sync_dirty_buffer(sbh); |
4066 | if (error) | |
4067 | return error; | |
4068 | ||
4069 | error = buffer_write_io_error(sbh); | |
4070 | if (error) { | |
b31e1552 ES |
4071 | ext4_msg(sb, KERN_ERR, "I/O error while writing " |
4072 | "superblock"); | |
914258bf TT |
4073 | clear_buffer_write_io_error(sbh); |
4074 | set_buffer_uptodate(sbh); | |
4075 | } | |
4076 | } | |
c4be0c1d | 4077 | return error; |
ac27a0ec DK |
4078 | } |
4079 | ||
ac27a0ec DK |
4080 | /* |
4081 | * Have we just finished recovery? If so, and if we are mounting (or | |
4082 | * remounting) the filesystem readonly, then we will end up with a | |
4083 | * consistent fs on disk. Record that fact. | |
4084 | */ | |
2b2d6d01 TT |
4085 | static void ext4_mark_recovery_complete(struct super_block *sb, |
4086 | struct ext4_super_block *es) | |
ac27a0ec | 4087 | { |
617ba13b | 4088 | journal_t *journal = EXT4_SB(sb)->s_journal; |
ac27a0ec | 4089 | |
0390131b FM |
4090 | if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { |
4091 | BUG_ON(journal != NULL); | |
4092 | return; | |
4093 | } | |
dab291af | 4094 | jbd2_journal_lock_updates(journal); |
7ffe1ea8 HK |
4095 | if (jbd2_journal_flush(journal) < 0) |
4096 | goto out; | |
4097 | ||
617ba13b | 4098 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) && |
ac27a0ec | 4099 | sb->s_flags & MS_RDONLY) { |
617ba13b | 4100 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); |
e2d67052 | 4101 | ext4_commit_super(sb, 1); |
ac27a0ec | 4102 | } |
7ffe1ea8 HK |
4103 | |
4104 | out: | |
dab291af | 4105 | jbd2_journal_unlock_updates(journal); |
ac27a0ec DK |
4106 | } |
4107 | ||
4108 | /* | |
4109 | * If we are mounting (or read-write remounting) a filesystem whose journal | |
4110 | * has recorded an error from a previous lifetime, move that error to the | |
4111 | * main filesystem now. | |
4112 | */ | |
2b2d6d01 TT |
4113 | static void ext4_clear_journal_err(struct super_block *sb, |
4114 | struct ext4_super_block *es) | |
ac27a0ec DK |
4115 | { |
4116 | journal_t *journal; | |
4117 | int j_errno; | |
4118 | const char *errstr; | |
4119 | ||
0390131b FM |
4120 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
4121 | ||
617ba13b | 4122 | journal = EXT4_SB(sb)->s_journal; |
ac27a0ec DK |
4123 | |
4124 | /* | |
4125 | * Now check for any error status which may have been recorded in the | |
617ba13b | 4126 | * journal by a prior ext4_error() or ext4_abort() |
ac27a0ec DK |
4127 | */ |
4128 | ||
dab291af | 4129 | j_errno = jbd2_journal_errno(journal); |
ac27a0ec DK |
4130 | if (j_errno) { |
4131 | char nbuf[16]; | |
4132 | ||
617ba13b | 4133 | errstr = ext4_decode_error(sb, j_errno, nbuf); |
12062ddd | 4134 | ext4_warning(sb, "Filesystem error recorded " |
ac27a0ec | 4135 | "from previous mount: %s", errstr); |
12062ddd | 4136 | ext4_warning(sb, "Marking fs in need of filesystem check."); |
ac27a0ec | 4137 | |
617ba13b MC |
4138 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; |
4139 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
e2d67052 | 4140 | ext4_commit_super(sb, 1); |
ac27a0ec | 4141 | |
dab291af | 4142 | jbd2_journal_clear_err(journal); |
ac27a0ec DK |
4143 | } |
4144 | } | |
4145 | ||
4146 | /* | |
4147 | * Force the running and committing transactions to commit, | |
4148 | * and wait on the commit. | |
4149 | */ | |
617ba13b | 4150 | int ext4_force_commit(struct super_block *sb) |
ac27a0ec DK |
4151 | { |
4152 | journal_t *journal; | |
0390131b | 4153 | int ret = 0; |
ac27a0ec DK |
4154 | |
4155 | if (sb->s_flags & MS_RDONLY) | |
4156 | return 0; | |
4157 | ||
617ba13b | 4158 | journal = EXT4_SB(sb)->s_journal; |
6b0310fb | 4159 | if (journal) { |
437f88cc | 4160 | vfs_check_frozen(sb, SB_FREEZE_TRANS); |
0390131b | 4161 | ret = ext4_journal_force_commit(journal); |
6b0310fb | 4162 | } |
0390131b | 4163 | |
ac27a0ec DK |
4164 | return ret; |
4165 | } | |
4166 | ||
2b2d6d01 | 4167 | static void ext4_write_super(struct super_block *sb) |
ac27a0ec | 4168 | { |
ebc1ac16 | 4169 | lock_super(sb); |
9ca92389 | 4170 | ext4_commit_super(sb, 1); |
ebc1ac16 | 4171 | unlock_super(sb); |
ac27a0ec DK |
4172 | } |
4173 | ||
617ba13b | 4174 | static int ext4_sync_fs(struct super_block *sb, int wait) |
ac27a0ec | 4175 | { |
14ce0cb4 | 4176 | int ret = 0; |
9eddacf9 | 4177 | tid_t target; |
8d5d02e6 | 4178 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 4179 | |
9bffad1e | 4180 | trace_ext4_sync_fs(sb, wait); |
8d5d02e6 MC |
4181 | flush_workqueue(sbi->dio_unwritten_wq); |
4182 | if (jbd2_journal_start_commit(sbi->s_journal, &target)) { | |
9ca92389 | 4183 | if (wait) |
8d5d02e6 | 4184 | jbd2_log_wait_commit(sbi->s_journal, target); |
0390131b | 4185 | } |
14ce0cb4 | 4186 | return ret; |
ac27a0ec DK |
4187 | } |
4188 | ||
4189 | /* | |
4190 | * LVM calls this function before a (read-only) snapshot is created. This | |
4191 | * gives us a chance to flush the journal completely and mark the fs clean. | |
be4f27d3 YY |
4192 | * |
4193 | * Note that only this function cannot bring a filesystem to be in a clean | |
4194 | * state independently, because ext4 prevents a new handle from being started | |
4195 | * by @sb->s_frozen, which stays in an upper layer. It thus needs help from | |
4196 | * the upper layer. | |
ac27a0ec | 4197 | */ |
c4be0c1d | 4198 | static int ext4_freeze(struct super_block *sb) |
ac27a0ec | 4199 | { |
c4be0c1d TS |
4200 | int error = 0; |
4201 | journal_t *journal; | |
ac27a0ec | 4202 | |
9ca92389 TT |
4203 | if (sb->s_flags & MS_RDONLY) |
4204 | return 0; | |
ac27a0ec | 4205 | |
9ca92389 | 4206 | journal = EXT4_SB(sb)->s_journal; |
7ffe1ea8 | 4207 | |
9ca92389 TT |
4208 | /* Now we set up the journal barrier. */ |
4209 | jbd2_journal_lock_updates(journal); | |
ac27a0ec | 4210 | |
9ca92389 TT |
4211 | /* |
4212 | * Don't clear the needs_recovery flag if we failed to flush | |
4213 | * the journal. | |
4214 | */ | |
4215 | error = jbd2_journal_flush(journal); | |
6b0310fb ES |
4216 | if (error < 0) |
4217 | goto out; | |
9ca92389 TT |
4218 | |
4219 | /* Journal blocked and flushed, clear needs_recovery flag. */ | |
4220 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
4221 | error = ext4_commit_super(sb, 1); | |
6b0310fb ES |
4222 | out: |
4223 | /* we rely on s_frozen to stop further updates */ | |
4224 | jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); | |
4225 | return error; | |
ac27a0ec DK |
4226 | } |
4227 | ||
4228 | /* | |
4229 | * Called by LVM after the snapshot is done. We need to reset the RECOVER | |
4230 | * flag here, even though the filesystem is not technically dirty yet. | |
4231 | */ | |
c4be0c1d | 4232 | static int ext4_unfreeze(struct super_block *sb) |
ac27a0ec | 4233 | { |
9ca92389 TT |
4234 | if (sb->s_flags & MS_RDONLY) |
4235 | return 0; | |
4236 | ||
4237 | lock_super(sb); | |
4238 | /* Reset the needs_recovery flag before the fs is unlocked. */ | |
4239 | EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
4240 | ext4_commit_super(sb, 1); | |
4241 | unlock_super(sb); | |
c4be0c1d | 4242 | return 0; |
ac27a0ec DK |
4243 | } |
4244 | ||
673c6100 TT |
4245 | /* |
4246 | * Structure to save mount options for ext4_remount's benefit | |
4247 | */ | |
4248 | struct ext4_mount_options { | |
4249 | unsigned long s_mount_opt; | |
a2595b8a | 4250 | unsigned long s_mount_opt2; |
673c6100 TT |
4251 | uid_t s_resuid; |
4252 | gid_t s_resgid; | |
4253 | unsigned long s_commit_interval; | |
4254 | u32 s_min_batch_time, s_max_batch_time; | |
4255 | #ifdef CONFIG_QUOTA | |
4256 | int s_jquota_fmt; | |
4257 | char *s_qf_names[MAXQUOTAS]; | |
4258 | #endif | |
4259 | }; | |
4260 | ||
2b2d6d01 | 4261 | static int ext4_remount(struct super_block *sb, int *flags, char *data) |
ac27a0ec | 4262 | { |
2b2d6d01 | 4263 | struct ext4_super_block *es; |
617ba13b MC |
4264 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
4265 | ext4_fsblk_t n_blocks_count = 0; | |
ac27a0ec | 4266 | unsigned long old_sb_flags; |
617ba13b | 4267 | struct ext4_mount_options old_opts; |
c79d967d | 4268 | int enable_quota = 0; |
8a266467 | 4269 | ext4_group_t g; |
b3881f74 | 4270 | unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; |
ac27a0ec DK |
4271 | int err; |
4272 | #ifdef CONFIG_QUOTA | |
4273 | int i; | |
4274 | #endif | |
d4c402d9 | 4275 | char *orig_data = kstrdup(data, GFP_KERNEL); |
ac27a0ec DK |
4276 | |
4277 | /* Store the original options */ | |
bbd6851a | 4278 | lock_super(sb); |
ac27a0ec DK |
4279 | old_sb_flags = sb->s_flags; |
4280 | old_opts.s_mount_opt = sbi->s_mount_opt; | |
a2595b8a | 4281 | old_opts.s_mount_opt2 = sbi->s_mount_opt2; |
ac27a0ec DK |
4282 | old_opts.s_resuid = sbi->s_resuid; |
4283 | old_opts.s_resgid = sbi->s_resgid; | |
4284 | old_opts.s_commit_interval = sbi->s_commit_interval; | |
30773840 TT |
4285 | old_opts.s_min_batch_time = sbi->s_min_batch_time; |
4286 | old_opts.s_max_batch_time = sbi->s_max_batch_time; | |
ac27a0ec DK |
4287 | #ifdef CONFIG_QUOTA |
4288 | old_opts.s_jquota_fmt = sbi->s_jquota_fmt; | |
4289 | for (i = 0; i < MAXQUOTAS; i++) | |
4290 | old_opts.s_qf_names[i] = sbi->s_qf_names[i]; | |
4291 | #endif | |
b3881f74 TT |
4292 | if (sbi->s_journal && sbi->s_journal->j_task->io_context) |
4293 | journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; | |
ac27a0ec DK |
4294 | |
4295 | /* | |
4296 | * Allow the "check" option to be passed as a remount option. | |
4297 | */ | |
b3881f74 TT |
4298 | if (!parse_options(data, sb, NULL, &journal_ioprio, |
4299 | &n_blocks_count, 1)) { | |
ac27a0ec DK |
4300 | err = -EINVAL; |
4301 | goto restore_opts; | |
4302 | } | |
4303 | ||
4ab2f15b | 4304 | if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) |
c67d859e | 4305 | ext4_abort(sb, "Abort forced by user"); |
ac27a0ec DK |
4306 | |
4307 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
482a7425 | 4308 | (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); |
ac27a0ec DK |
4309 | |
4310 | es = sbi->s_es; | |
4311 | ||
b3881f74 | 4312 | if (sbi->s_journal) { |
0390131b | 4313 | ext4_init_journal_params(sb, sbi->s_journal); |
b3881f74 TT |
4314 | set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); |
4315 | } | |
ac27a0ec DK |
4316 | |
4317 | if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) || | |
bd81d8ee | 4318 | n_blocks_count > ext4_blocks_count(es)) { |
4ab2f15b | 4319 | if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { |
ac27a0ec DK |
4320 | err = -EROFS; |
4321 | goto restore_opts; | |
4322 | } | |
4323 | ||
4324 | if (*flags & MS_RDONLY) { | |
0f0dd62f CH |
4325 | err = dquot_suspend(sb, -1); |
4326 | if (err < 0) | |
c79d967d | 4327 | goto restore_opts; |
c79d967d | 4328 | |
ac27a0ec DK |
4329 | /* |
4330 | * First of all, the unconditional stuff we have to do | |
4331 | * to disable replay of the journal when we next remount | |
4332 | */ | |
4333 | sb->s_flags |= MS_RDONLY; | |
4334 | ||
4335 | /* | |
4336 | * OK, test if we are remounting a valid rw partition | |
4337 | * readonly, and if so set the rdonly flag and then | |
4338 | * mark the partition as valid again. | |
4339 | */ | |
617ba13b MC |
4340 | if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && |
4341 | (sbi->s_mount_state & EXT4_VALID_FS)) | |
ac27a0ec DK |
4342 | es->s_state = cpu_to_le16(sbi->s_mount_state); |
4343 | ||
a63c9eb2 | 4344 | if (sbi->s_journal) |
0390131b | 4345 | ext4_mark_recovery_complete(sb, es); |
ac27a0ec | 4346 | } else { |
a13fb1a4 ES |
4347 | /* Make sure we can mount this feature set readwrite */ |
4348 | if (!ext4_feature_set_ok(sb, 0)) { | |
ac27a0ec DK |
4349 | err = -EROFS; |
4350 | goto restore_opts; | |
4351 | } | |
8a266467 TT |
4352 | /* |
4353 | * Make sure the group descriptor checksums | |
0b8e58a1 | 4354 | * are sane. If they aren't, refuse to remount r/w. |
8a266467 TT |
4355 | */ |
4356 | for (g = 0; g < sbi->s_groups_count; g++) { | |
4357 | struct ext4_group_desc *gdp = | |
4358 | ext4_get_group_desc(sb, g, NULL); | |
4359 | ||
4360 | if (!ext4_group_desc_csum_verify(sbi, g, gdp)) { | |
b31e1552 ES |
4361 | ext4_msg(sb, KERN_ERR, |
4362 | "ext4_remount: Checksum for group %u failed (%u!=%u)", | |
8a266467 TT |
4363 | g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)), |
4364 | le16_to_cpu(gdp->bg_checksum)); | |
4365 | err = -EINVAL; | |
4366 | goto restore_opts; | |
4367 | } | |
4368 | } | |
4369 | ||
ead6596b ES |
4370 | /* |
4371 | * If we have an unprocessed orphan list hanging | |
4372 | * around from a previously readonly bdev mount, | |
4373 | * require a full umount/remount for now. | |
4374 | */ | |
4375 | if (es->s_last_orphan) { | |
b31e1552 | 4376 | ext4_msg(sb, KERN_WARNING, "Couldn't " |
ead6596b ES |
4377 | "remount RDWR because of unprocessed " |
4378 | "orphan inode list. Please " | |
b31e1552 | 4379 | "umount/remount instead"); |
ead6596b ES |
4380 | err = -EINVAL; |
4381 | goto restore_opts; | |
4382 | } | |
4383 | ||
ac27a0ec DK |
4384 | /* |
4385 | * Mounting a RDONLY partition read-write, so reread | |
4386 | * and store the current valid flag. (It may have | |
4387 | * been changed by e2fsck since we originally mounted | |
4388 | * the partition.) | |
4389 | */ | |
0390131b FM |
4390 | if (sbi->s_journal) |
4391 | ext4_clear_journal_err(sb, es); | |
ac27a0ec | 4392 | sbi->s_mount_state = le16_to_cpu(es->s_state); |
617ba13b | 4393 | if ((err = ext4_group_extend(sb, es, n_blocks_count))) |
ac27a0ec | 4394 | goto restore_opts; |
2b2d6d01 | 4395 | if (!ext4_setup_super(sb, es, 0)) |
ac27a0ec | 4396 | sb->s_flags &= ~MS_RDONLY; |
c79d967d | 4397 | enable_quota = 1; |
ac27a0ec DK |
4398 | } |
4399 | } | |
bfff6873 LC |
4400 | |
4401 | /* | |
4402 | * Reinitialize lazy itable initialization thread based on | |
4403 | * current settings | |
4404 | */ | |
4405 | if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE)) | |
4406 | ext4_unregister_li_request(sb); | |
4407 | else { | |
4408 | ext4_group_t first_not_zeroed; | |
4409 | first_not_zeroed = ext4_has_uninit_itable(sb); | |
4410 | ext4_register_li_request(sb, first_not_zeroed); | |
4411 | } | |
4412 | ||
6fd058f7 | 4413 | ext4_setup_system_zone(sb); |
0390131b | 4414 | if (sbi->s_journal == NULL) |
e2d67052 | 4415 | ext4_commit_super(sb, 1); |
0390131b | 4416 | |
ac27a0ec DK |
4417 | #ifdef CONFIG_QUOTA |
4418 | /* Release old quota file names */ | |
4419 | for (i = 0; i < MAXQUOTAS; i++) | |
4420 | if (old_opts.s_qf_names[i] && | |
4421 | old_opts.s_qf_names[i] != sbi->s_qf_names[i]) | |
4422 | kfree(old_opts.s_qf_names[i]); | |
4423 | #endif | |
bbd6851a | 4424 | unlock_super(sb); |
c79d967d | 4425 | if (enable_quota) |
0f0dd62f | 4426 | dquot_resume(sb, -1); |
d4c402d9 CW |
4427 | |
4428 | ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); | |
4429 | kfree(orig_data); | |
ac27a0ec | 4430 | return 0; |
0b8e58a1 | 4431 | |
ac27a0ec DK |
4432 | restore_opts: |
4433 | sb->s_flags = old_sb_flags; | |
4434 | sbi->s_mount_opt = old_opts.s_mount_opt; | |
a2595b8a | 4435 | sbi->s_mount_opt2 = old_opts.s_mount_opt2; |
ac27a0ec DK |
4436 | sbi->s_resuid = old_opts.s_resuid; |
4437 | sbi->s_resgid = old_opts.s_resgid; | |
4438 | sbi->s_commit_interval = old_opts.s_commit_interval; | |
30773840 TT |
4439 | sbi->s_min_batch_time = old_opts.s_min_batch_time; |
4440 | sbi->s_max_batch_time = old_opts.s_max_batch_time; | |
ac27a0ec DK |
4441 | #ifdef CONFIG_QUOTA |
4442 | sbi->s_jquota_fmt = old_opts.s_jquota_fmt; | |
4443 | for (i = 0; i < MAXQUOTAS; i++) { | |
4444 | if (sbi->s_qf_names[i] && | |
4445 | old_opts.s_qf_names[i] != sbi->s_qf_names[i]) | |
4446 | kfree(sbi->s_qf_names[i]); | |
4447 | sbi->s_qf_names[i] = old_opts.s_qf_names[i]; | |
4448 | } | |
4449 | #endif | |
bbd6851a | 4450 | unlock_super(sb); |
d4c402d9 | 4451 | kfree(orig_data); |
ac27a0ec DK |
4452 | return err; |
4453 | } | |
4454 | ||
2b2d6d01 | 4455 | static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) |
ac27a0ec DK |
4456 | { |
4457 | struct super_block *sb = dentry->d_sb; | |
617ba13b MC |
4458 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
4459 | struct ext4_super_block *es = sbi->s_es; | |
960cc398 | 4460 | u64 fsid; |
ac27a0ec | 4461 | |
5e70030d BP |
4462 | if (test_opt(sb, MINIX_DF)) { |
4463 | sbi->s_overhead_last = 0; | |
6bc9feff | 4464 | } else if (sbi->s_blocks_last != ext4_blocks_count(es)) { |
8df9675f | 4465 | ext4_group_t i, ngroups = ext4_get_groups_count(sb); |
5e70030d | 4466 | ext4_fsblk_t overhead = 0; |
ac27a0ec DK |
4467 | |
4468 | /* | |
5e70030d BP |
4469 | * Compute the overhead (FS structures). This is constant |
4470 | * for a given filesystem unless the number of block groups | |
4471 | * changes so we cache the previous value until it does. | |
ac27a0ec DK |
4472 | */ |
4473 | ||
4474 | /* | |
4475 | * All of the blocks before first_data_block are | |
4476 | * overhead | |
4477 | */ | |
4478 | overhead = le32_to_cpu(es->s_first_data_block); | |
4479 | ||
4480 | /* | |
4481 | * Add the overhead attributed to the superblock and | |
4482 | * block group descriptors. If the sparse superblocks | |
4483 | * feature is turned on, then not all groups have this. | |
4484 | */ | |
4485 | for (i = 0; i < ngroups; i++) { | |
617ba13b MC |
4486 | overhead += ext4_bg_has_super(sb, i) + |
4487 | ext4_bg_num_gdb(sb, i); | |
ac27a0ec DK |
4488 | cond_resched(); |
4489 | } | |
4490 | ||
4491 | /* | |
4492 | * Every block group has an inode bitmap, a block | |
4493 | * bitmap, and an inode table. | |
4494 | */ | |
5e70030d BP |
4495 | overhead += ngroups * (2 + sbi->s_itb_per_group); |
4496 | sbi->s_overhead_last = overhead; | |
4497 | smp_wmb(); | |
6bc9feff | 4498 | sbi->s_blocks_last = ext4_blocks_count(es); |
ac27a0ec DK |
4499 | } |
4500 | ||
617ba13b | 4501 | buf->f_type = EXT4_SUPER_MAGIC; |
ac27a0ec | 4502 | buf->f_bsize = sb->s_blocksize; |
5e70030d | 4503 | buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last; |
6bc6e63f AK |
4504 | buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) - |
4505 | percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter); | |
bd81d8ee LV |
4506 | buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es); |
4507 | if (buf->f_bfree < ext4_r_blocks_count(es)) | |
ac27a0ec DK |
4508 | buf->f_bavail = 0; |
4509 | buf->f_files = le32_to_cpu(es->s_inodes_count); | |
52d9f3b4 | 4510 | buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); |
617ba13b | 4511 | buf->f_namelen = EXT4_NAME_LEN; |
960cc398 PE |
4512 | fsid = le64_to_cpup((void *)es->s_uuid) ^ |
4513 | le64_to_cpup((void *)es->s_uuid + sizeof(u64)); | |
4514 | buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; | |
4515 | buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; | |
0b8e58a1 | 4516 | |
ac27a0ec DK |
4517 | return 0; |
4518 | } | |
4519 | ||
0b8e58a1 AD |
4520 | /* Helper function for writing quotas on sync - we need to start transaction |
4521 | * before quota file is locked for write. Otherwise the are possible deadlocks: | |
ac27a0ec | 4522 | * Process 1 Process 2 |
617ba13b | 4523 | * ext4_create() quota_sync() |
a269eb18 | 4524 | * jbd2_journal_start() write_dquot() |
871a2931 | 4525 | * dquot_initialize() down(dqio_mutex) |
dab291af | 4526 | * down(dqio_mutex) jbd2_journal_start() |
ac27a0ec DK |
4527 | * |
4528 | */ | |
4529 | ||
4530 | #ifdef CONFIG_QUOTA | |
4531 | ||
4532 | static inline struct inode *dquot_to_inode(struct dquot *dquot) | |
4533 | { | |
4534 | return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type]; | |
4535 | } | |
4536 | ||
617ba13b | 4537 | static int ext4_write_dquot(struct dquot *dquot) |
ac27a0ec DK |
4538 | { |
4539 | int ret, err; | |
4540 | handle_t *handle; | |
4541 | struct inode *inode; | |
4542 | ||
4543 | inode = dquot_to_inode(dquot); | |
617ba13b | 4544 | handle = ext4_journal_start(inode, |
0b8e58a1 | 4545 | EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); |
ac27a0ec DK |
4546 | if (IS_ERR(handle)) |
4547 | return PTR_ERR(handle); | |
4548 | ret = dquot_commit(dquot); | |
617ba13b | 4549 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
4550 | if (!ret) |
4551 | ret = err; | |
4552 | return ret; | |
4553 | } | |
4554 | ||
617ba13b | 4555 | static int ext4_acquire_dquot(struct dquot *dquot) |
ac27a0ec DK |
4556 | { |
4557 | int ret, err; | |
4558 | handle_t *handle; | |
4559 | ||
617ba13b | 4560 | handle = ext4_journal_start(dquot_to_inode(dquot), |
0b8e58a1 | 4561 | EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); |
ac27a0ec DK |
4562 | if (IS_ERR(handle)) |
4563 | return PTR_ERR(handle); | |
4564 | ret = dquot_acquire(dquot); | |
617ba13b | 4565 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
4566 | if (!ret) |
4567 | ret = err; | |
4568 | return ret; | |
4569 | } | |
4570 | ||
617ba13b | 4571 | static int ext4_release_dquot(struct dquot *dquot) |
ac27a0ec DK |
4572 | { |
4573 | int ret, err; | |
4574 | handle_t *handle; | |
4575 | ||
617ba13b | 4576 | handle = ext4_journal_start(dquot_to_inode(dquot), |
0b8e58a1 | 4577 | EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); |
9c3013e9 JK |
4578 | if (IS_ERR(handle)) { |
4579 | /* Release dquot anyway to avoid endless cycle in dqput() */ | |
4580 | dquot_release(dquot); | |
ac27a0ec | 4581 | return PTR_ERR(handle); |
9c3013e9 | 4582 | } |
ac27a0ec | 4583 | ret = dquot_release(dquot); |
617ba13b | 4584 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
4585 | if (!ret) |
4586 | ret = err; | |
4587 | return ret; | |
4588 | } | |
4589 | ||
617ba13b | 4590 | static int ext4_mark_dquot_dirty(struct dquot *dquot) |
ac27a0ec | 4591 | { |
2c8be6b2 | 4592 | /* Are we journaling quotas? */ |
617ba13b MC |
4593 | if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] || |
4594 | EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) { | |
ac27a0ec | 4595 | dquot_mark_dquot_dirty(dquot); |
617ba13b | 4596 | return ext4_write_dquot(dquot); |
ac27a0ec DK |
4597 | } else { |
4598 | return dquot_mark_dquot_dirty(dquot); | |
4599 | } | |
4600 | } | |
4601 | ||
617ba13b | 4602 | static int ext4_write_info(struct super_block *sb, int type) |
ac27a0ec DK |
4603 | { |
4604 | int ret, err; | |
4605 | handle_t *handle; | |
4606 | ||
4607 | /* Data block + inode block */ | |
617ba13b | 4608 | handle = ext4_journal_start(sb->s_root->d_inode, 2); |
ac27a0ec DK |
4609 | if (IS_ERR(handle)) |
4610 | return PTR_ERR(handle); | |
4611 | ret = dquot_commit_info(sb, type); | |
617ba13b | 4612 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
4613 | if (!ret) |
4614 | ret = err; | |
4615 | return ret; | |
4616 | } | |
4617 | ||
4618 | /* | |
4619 | * Turn on quotas during mount time - we need to find | |
4620 | * the quota file and such... | |
4621 | */ | |
617ba13b | 4622 | static int ext4_quota_on_mount(struct super_block *sb, int type) |
ac27a0ec | 4623 | { |
287a8095 CH |
4624 | return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type], |
4625 | EXT4_SB(sb)->s_jquota_fmt, type); | |
ac27a0ec DK |
4626 | } |
4627 | ||
4628 | /* | |
4629 | * Standard function to be called on quota_on | |
4630 | */ | |
617ba13b | 4631 | static int ext4_quota_on(struct super_block *sb, int type, int format_id, |
f00c9e44 | 4632 | struct path *path) |
ac27a0ec DK |
4633 | { |
4634 | int err; | |
ac27a0ec DK |
4635 | |
4636 | if (!test_opt(sb, QUOTA)) | |
4637 | return -EINVAL; | |
0623543b | 4638 | |
ac27a0ec | 4639 | /* Quotafile not on the same filesystem? */ |
f00c9e44 | 4640 | if (path->mnt->mnt_sb != sb) |
ac27a0ec | 4641 | return -EXDEV; |
0623543b JK |
4642 | /* Journaling quota? */ |
4643 | if (EXT4_SB(sb)->s_qf_names[type]) { | |
2b2d6d01 | 4644 | /* Quotafile not in fs root? */ |
f00c9e44 | 4645 | if (path->dentry->d_parent != sb->s_root) |
b31e1552 ES |
4646 | ext4_msg(sb, KERN_WARNING, |
4647 | "Quota file not on filesystem root. " | |
4648 | "Journaled quota will not work"); | |
2b2d6d01 | 4649 | } |
0623543b JK |
4650 | |
4651 | /* | |
4652 | * When we journal data on quota file, we have to flush journal to see | |
4653 | * all updates to the file when we bypass pagecache... | |
4654 | */ | |
0390131b | 4655 | if (EXT4_SB(sb)->s_journal && |
f00c9e44 | 4656 | ext4_should_journal_data(path->dentry->d_inode)) { |
0623543b JK |
4657 | /* |
4658 | * We don't need to lock updates but journal_flush() could | |
4659 | * otherwise be livelocked... | |
4660 | */ | |
4661 | jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); | |
7ffe1ea8 | 4662 | err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); |
0623543b | 4663 | jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); |
f00c9e44 | 4664 | if (err) |
7ffe1ea8 | 4665 | return err; |
0623543b JK |
4666 | } |
4667 | ||
f00c9e44 | 4668 | return dquot_quota_on(sb, type, format_id, path); |
ac27a0ec DK |
4669 | } |
4670 | ||
ca0e05e4 DM |
4671 | static int ext4_quota_off(struct super_block *sb, int type) |
4672 | { | |
21f97697 JK |
4673 | struct inode *inode = sb_dqopt(sb)->files[type]; |
4674 | handle_t *handle; | |
4675 | ||
87009d86 DM |
4676 | /* Force all delayed allocation blocks to be allocated. |
4677 | * Caller already holds s_umount sem */ | |
4678 | if (test_opt(sb, DELALLOC)) | |
ca0e05e4 | 4679 | sync_filesystem(sb); |
ca0e05e4 | 4680 | |
0b268590 AG |
4681 | if (!inode) |
4682 | goto out; | |
4683 | ||
21f97697 JK |
4684 | /* Update modification times of quota files when userspace can |
4685 | * start looking at them */ | |
4686 | handle = ext4_journal_start(inode, 1); | |
4687 | if (IS_ERR(handle)) | |
4688 | goto out; | |
4689 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; | |
4690 | ext4_mark_inode_dirty(handle, inode); | |
4691 | ext4_journal_stop(handle); | |
4692 | ||
4693 | out: | |
ca0e05e4 DM |
4694 | return dquot_quota_off(sb, type); |
4695 | } | |
4696 | ||
ac27a0ec DK |
4697 | /* Read data from quotafile - avoid pagecache and such because we cannot afford |
4698 | * acquiring the locks... As quota files are never truncated and quota code | |
25985edc | 4699 | * itself serializes the operations (and no one else should touch the files) |
ac27a0ec | 4700 | * we don't have to be afraid of races */ |
617ba13b | 4701 | static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, |
ac27a0ec DK |
4702 | size_t len, loff_t off) |
4703 | { | |
4704 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
725d26d3 | 4705 | ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); |
ac27a0ec DK |
4706 | int err = 0; |
4707 | int offset = off & (sb->s_blocksize - 1); | |
4708 | int tocopy; | |
4709 | size_t toread; | |
4710 | struct buffer_head *bh; | |
4711 | loff_t i_size = i_size_read(inode); | |
4712 | ||
4713 | if (off > i_size) | |
4714 | return 0; | |
4715 | if (off+len > i_size) | |
4716 | len = i_size-off; | |
4717 | toread = len; | |
4718 | while (toread > 0) { | |
4719 | tocopy = sb->s_blocksize - offset < toread ? | |
4720 | sb->s_blocksize - offset : toread; | |
617ba13b | 4721 | bh = ext4_bread(NULL, inode, blk, 0, &err); |
ac27a0ec DK |
4722 | if (err) |
4723 | return err; | |
4724 | if (!bh) /* A hole? */ | |
4725 | memset(data, 0, tocopy); | |
4726 | else | |
4727 | memcpy(data, bh->b_data+offset, tocopy); | |
4728 | brelse(bh); | |
4729 | offset = 0; | |
4730 | toread -= tocopy; | |
4731 | data += tocopy; | |
4732 | blk++; | |
4733 | } | |
4734 | return len; | |
4735 | } | |
4736 | ||
4737 | /* Write to quotafile (we know the transaction is already started and has | |
4738 | * enough credits) */ | |
617ba13b | 4739 | static ssize_t ext4_quota_write(struct super_block *sb, int type, |
ac27a0ec DK |
4740 | const char *data, size_t len, loff_t off) |
4741 | { | |
4742 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
725d26d3 | 4743 | ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); |
ac27a0ec DK |
4744 | int err = 0; |
4745 | int offset = off & (sb->s_blocksize - 1); | |
ac27a0ec DK |
4746 | struct buffer_head *bh; |
4747 | handle_t *handle = journal_current_handle(); | |
4748 | ||
0390131b | 4749 | if (EXT4_SB(sb)->s_journal && !handle) { |
b31e1552 ES |
4750 | ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" |
4751 | " cancelled because transaction is not started", | |
9c3013e9 JK |
4752 | (unsigned long long)off, (unsigned long long)len); |
4753 | return -EIO; | |
4754 | } | |
67eeb568 DM |
4755 | /* |
4756 | * Since we account only one data block in transaction credits, | |
4757 | * then it is impossible to cross a block boundary. | |
4758 | */ | |
4759 | if (sb->s_blocksize - offset < len) { | |
4760 | ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" | |
4761 | " cancelled because not block aligned", | |
4762 | (unsigned long long)off, (unsigned long long)len); | |
4763 | return -EIO; | |
4764 | } | |
4765 | ||
ac27a0ec | 4766 | mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA); |
67eeb568 DM |
4767 | bh = ext4_bread(handle, inode, blk, 1, &err); |
4768 | if (!bh) | |
4769 | goto out; | |
62d2b5f2 JK |
4770 | err = ext4_journal_get_write_access(handle, bh); |
4771 | if (err) { | |
4772 | brelse(bh); | |
4773 | goto out; | |
ac27a0ec | 4774 | } |
67eeb568 DM |
4775 | lock_buffer(bh); |
4776 | memcpy(bh->b_data+offset, data, len); | |
4777 | flush_dcache_page(bh->b_page); | |
4778 | unlock_buffer(bh); | |
62d2b5f2 | 4779 | err = ext4_handle_dirty_metadata(handle, NULL, bh); |
67eeb568 | 4780 | brelse(bh); |
ac27a0ec | 4781 | out: |
67eeb568 | 4782 | if (err) { |
4d04e4fb | 4783 | mutex_unlock(&inode->i_mutex); |
ac27a0ec | 4784 | return err; |
4d04e4fb | 4785 | } |
67eeb568 DM |
4786 | if (inode->i_size < off + len) { |
4787 | i_size_write(inode, off + len); | |
617ba13b | 4788 | EXT4_I(inode)->i_disksize = inode->i_size; |
21f97697 | 4789 | ext4_mark_inode_dirty(handle, inode); |
ac27a0ec | 4790 | } |
ac27a0ec | 4791 | mutex_unlock(&inode->i_mutex); |
67eeb568 | 4792 | return len; |
ac27a0ec DK |
4793 | } |
4794 | ||
4795 | #endif | |
4796 | ||
152a0836 AV |
4797 | static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, |
4798 | const char *dev_name, void *data) | |
ac27a0ec | 4799 | { |
152a0836 | 4800 | return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); |
ac27a0ec DK |
4801 | } |
4802 | ||
37f328eb | 4803 | #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) |
24b58424 TT |
4804 | static inline void register_as_ext2(void) |
4805 | { | |
4806 | int err = register_filesystem(&ext2_fs_type); | |
4807 | if (err) | |
4808 | printk(KERN_WARNING | |
4809 | "EXT4-fs: Unable to register as ext2 (%d)\n", err); | |
4810 | } | |
4811 | ||
4812 | static inline void unregister_as_ext2(void) | |
4813 | { | |
4814 | unregister_filesystem(&ext2_fs_type); | |
4815 | } | |
2035e776 TT |
4816 | |
4817 | static inline int ext2_feature_set_ok(struct super_block *sb) | |
4818 | { | |
4819 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP)) | |
4820 | return 0; | |
4821 | if (sb->s_flags & MS_RDONLY) | |
4822 | return 1; | |
4823 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP)) | |
4824 | return 0; | |
4825 | return 1; | |
4826 | } | |
51b7e3c9 | 4827 | MODULE_ALIAS("ext2"); |
24b58424 TT |
4828 | #else |
4829 | static inline void register_as_ext2(void) { } | |
4830 | static inline void unregister_as_ext2(void) { } | |
2035e776 | 4831 | static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } |
24b58424 TT |
4832 | #endif |
4833 | ||
37f328eb | 4834 | #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) |
24b58424 TT |
4835 | static inline void register_as_ext3(void) |
4836 | { | |
4837 | int err = register_filesystem(&ext3_fs_type); | |
4838 | if (err) | |
4839 | printk(KERN_WARNING | |
4840 | "EXT4-fs: Unable to register as ext3 (%d)\n", err); | |
4841 | } | |
4842 | ||
4843 | static inline void unregister_as_ext3(void) | |
4844 | { | |
4845 | unregister_filesystem(&ext3_fs_type); | |
4846 | } | |
2035e776 TT |
4847 | |
4848 | static inline int ext3_feature_set_ok(struct super_block *sb) | |
4849 | { | |
4850 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP)) | |
4851 | return 0; | |
4852 | if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) | |
4853 | return 0; | |
4854 | if (sb->s_flags & MS_RDONLY) | |
4855 | return 1; | |
4856 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP)) | |
4857 | return 0; | |
4858 | return 1; | |
4859 | } | |
51b7e3c9 | 4860 | MODULE_ALIAS("ext3"); |
24b58424 TT |
4861 | #else |
4862 | static inline void register_as_ext3(void) { } | |
4863 | static inline void unregister_as_ext3(void) { } | |
2035e776 | 4864 | static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; } |
24b58424 TT |
4865 | #endif |
4866 | ||
03010a33 TT |
4867 | static struct file_system_type ext4_fs_type = { |
4868 | .owner = THIS_MODULE, | |
4869 | .name = "ext4", | |
152a0836 | 4870 | .mount = ext4_mount, |
03010a33 TT |
4871 | .kill_sb = kill_block_super, |
4872 | .fs_flags = FS_REQUIRES_DEV, | |
4873 | }; | |
4874 | ||
8f021222 | 4875 | static int __init ext4_init_feat_adverts(void) |
857ac889 LC |
4876 | { |
4877 | struct ext4_features *ef; | |
4878 | int ret = -ENOMEM; | |
4879 | ||
4880 | ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL); | |
4881 | if (!ef) | |
4882 | goto out; | |
4883 | ||
4884 | ef->f_kobj.kset = ext4_kset; | |
4885 | init_completion(&ef->f_kobj_unregister); | |
4886 | ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL, | |
4887 | "features"); | |
4888 | if (ret) { | |
4889 | kfree(ef); | |
4890 | goto out; | |
4891 | } | |
4892 | ||
4893 | ext4_feat = ef; | |
4894 | ret = 0; | |
4895 | out: | |
4896 | return ret; | |
4897 | } | |
4898 | ||
8f021222 LC |
4899 | static void ext4_exit_feat_adverts(void) |
4900 | { | |
4901 | kobject_put(&ext4_feat->f_kobj); | |
4902 | wait_for_completion(&ext4_feat->f_kobj_unregister); | |
4903 | kfree(ext4_feat); | |
4904 | } | |
4905 | ||
e9e3bcec ES |
4906 | /* Shared across all ext4 file systems */ |
4907 | wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; | |
4908 | struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ]; | |
4909 | ||
5dabfc78 | 4910 | static int __init ext4_init_fs(void) |
ac27a0ec | 4911 | { |
e9e3bcec | 4912 | int i, err; |
c9de560d | 4913 | |
12e9b892 | 4914 | ext4_check_flag_values(); |
e9e3bcec ES |
4915 | |
4916 | for (i = 0; i < EXT4_WQ_HASH_SZ; i++) { | |
4917 | mutex_init(&ext4__aio_mutex[i]); | |
4918 | init_waitqueue_head(&ext4__ioend_wq[i]); | |
4919 | } | |
4920 | ||
5dabfc78 | 4921 | err = ext4_init_pageio(); |
6fd058f7 TT |
4922 | if (err) |
4923 | return err; | |
5dabfc78 | 4924 | err = ext4_init_system_zone(); |
bd2d0210 | 4925 | if (err) |
dd68314c | 4926 | goto out7; |
3197ebdb TT |
4927 | ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj); |
4928 | if (!ext4_kset) | |
dd68314c | 4929 | goto out6; |
9f6200bb | 4930 | ext4_proc_root = proc_mkdir("fs/ext4", NULL); |
dd68314c TT |
4931 | if (!ext4_proc_root) |
4932 | goto out5; | |
857ac889 LC |
4933 | |
4934 | err = ext4_init_feat_adverts(); | |
dd68314c TT |
4935 | if (err) |
4936 | goto out4; | |
857ac889 | 4937 | |
5dabfc78 | 4938 | err = ext4_init_mballoc(); |
ac27a0ec | 4939 | if (err) |
6fd058f7 | 4940 | goto out3; |
c9de560d | 4941 | |
5dabfc78 | 4942 | err = ext4_init_xattr(); |
c9de560d AT |
4943 | if (err) |
4944 | goto out2; | |
ac27a0ec DK |
4945 | err = init_inodecache(); |
4946 | if (err) | |
4947 | goto out1; | |
24b58424 | 4948 | register_as_ext3(); |
2035e776 | 4949 | register_as_ext2(); |
03010a33 | 4950 | err = register_filesystem(&ext4_fs_type); |
ac27a0ec DK |
4951 | if (err) |
4952 | goto out; | |
bfff6873 LC |
4953 | |
4954 | ext4_li_info = NULL; | |
4955 | mutex_init(&ext4_li_mtx); | |
ac27a0ec DK |
4956 | return 0; |
4957 | out: | |
24b58424 TT |
4958 | unregister_as_ext2(); |
4959 | unregister_as_ext3(); | |
ac27a0ec DK |
4960 | destroy_inodecache(); |
4961 | out1: | |
5dabfc78 | 4962 | ext4_exit_xattr(); |
c9de560d | 4963 | out2: |
5dabfc78 | 4964 | ext4_exit_mballoc(); |
6fd058f7 | 4965 | out3: |
8f021222 | 4966 | ext4_exit_feat_adverts(); |
dd68314c | 4967 | out4: |
6fd058f7 | 4968 | remove_proc_entry("fs/ext4", NULL); |
dd68314c | 4969 | out5: |
6fd058f7 | 4970 | kset_unregister(ext4_kset); |
dd68314c | 4971 | out6: |
5dabfc78 | 4972 | ext4_exit_system_zone(); |
dd68314c | 4973 | out7: |
5dabfc78 | 4974 | ext4_exit_pageio(); |
ac27a0ec DK |
4975 | return err; |
4976 | } | |
4977 | ||
5dabfc78 | 4978 | static void __exit ext4_exit_fs(void) |
ac27a0ec | 4979 | { |
bfff6873 | 4980 | ext4_destroy_lazyinit_thread(); |
24b58424 TT |
4981 | unregister_as_ext2(); |
4982 | unregister_as_ext3(); | |
03010a33 | 4983 | unregister_filesystem(&ext4_fs_type); |
ac27a0ec | 4984 | destroy_inodecache(); |
5dabfc78 TT |
4985 | ext4_exit_xattr(); |
4986 | ext4_exit_mballoc(); | |
8f021222 | 4987 | ext4_exit_feat_adverts(); |
9f6200bb | 4988 | remove_proc_entry("fs/ext4", NULL); |
3197ebdb | 4989 | kset_unregister(ext4_kset); |
5dabfc78 TT |
4990 | ext4_exit_system_zone(); |
4991 | ext4_exit_pageio(); | |
ac27a0ec DK |
4992 | } |
4993 | ||
4994 | MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); | |
83982b6f | 4995 | MODULE_DESCRIPTION("Fourth Extended Filesystem"); |
ac27a0ec | 4996 | MODULE_LICENSE("GPL"); |
5dabfc78 TT |
4997 | module_init(ext4_init_fs) |
4998 | module_exit(ext4_exit_fs) |