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> |
7abc52c2 | 41 | #include <linux/cleancache.h> |
ac27a0ec DK |
42 | #include <asm/uaccess.h> |
43 | ||
bfff6873 LC |
44 | #include <linux/kthread.h> |
45 | #include <linux/freezer.h> | |
46 | ||
3dcf5451 CH |
47 | #include "ext4.h" |
48 | #include "ext4_jbd2.h" | |
ac27a0ec DK |
49 | #include "xattr.h" |
50 | #include "acl.h" | |
3661d286 | 51 | #include "mballoc.h" |
ac27a0ec | 52 | |
9bffad1e TT |
53 | #define CREATE_TRACE_POINTS |
54 | #include <trace/events/ext4.h> | |
55 | ||
1f109d5a | 56 | static struct proc_dir_entry *ext4_proc_root; |
3197ebdb | 57 | static struct kset *ext4_kset; |
0b75a840 LC |
58 | static struct ext4_lazy_init *ext4_li_info; |
59 | static struct mutex ext4_li_mtx; | |
60 | static struct ext4_features *ext4_feat; | |
9f6200bb | 61 | |
617ba13b | 62 | static int ext4_load_journal(struct super_block *, struct ext4_super_block *, |
ac27a0ec | 63 | unsigned long journal_devnum); |
e2d67052 | 64 | static int ext4_commit_super(struct super_block *sb, int sync); |
2b2d6d01 TT |
65 | static void ext4_mark_recovery_complete(struct super_block *sb, |
66 | struct ext4_super_block *es); | |
67 | static void ext4_clear_journal_err(struct super_block *sb, | |
68 | struct ext4_super_block *es); | |
617ba13b | 69 | static int ext4_sync_fs(struct super_block *sb, int wait); |
2b2d6d01 | 70 | static const char *ext4_decode_error(struct super_block *sb, int errno, |
ac27a0ec | 71 | char nbuf[16]); |
2b2d6d01 TT |
72 | static int ext4_remount(struct super_block *sb, int *flags, char *data); |
73 | static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); | |
c4be0c1d | 74 | static int ext4_unfreeze(struct super_block *sb); |
2b2d6d01 | 75 | static void ext4_write_super(struct super_block *sb); |
c4be0c1d | 76 | static int ext4_freeze(struct super_block *sb); |
152a0836 AV |
77 | static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, |
78 | const char *dev_name, void *data); | |
2035e776 TT |
79 | static inline int ext2_feature_set_ok(struct super_block *sb); |
80 | static inline int ext3_feature_set_ok(struct super_block *sb); | |
d39195c3 | 81 | static int ext4_feature_set_ok(struct super_block *sb, int readonly); |
bfff6873 LC |
82 | static void ext4_destroy_lazyinit_thread(void); |
83 | static void ext4_unregister_li_request(struct super_block *sb); | |
8f1f7453 | 84 | static void ext4_clear_request_list(void); |
ac27a0ec | 85 | |
2035e776 TT |
86 | #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) |
87 | static struct file_system_type ext2_fs_type = { | |
88 | .owner = THIS_MODULE, | |
89 | .name = "ext2", | |
90 | .mount = ext4_mount, | |
91 | .kill_sb = kill_block_super, | |
92 | .fs_flags = FS_REQUIRES_DEV, | |
93 | }; | |
94 | #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) | |
95 | #else | |
96 | #define IS_EXT2_SB(sb) (0) | |
97 | #endif | |
98 | ||
99 | ||
ba69f9ab JK |
100 | #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) |
101 | static struct file_system_type ext3_fs_type = { | |
102 | .owner = THIS_MODULE, | |
103 | .name = "ext3", | |
152a0836 | 104 | .mount = ext4_mount, |
ba69f9ab JK |
105 | .kill_sb = kill_block_super, |
106 | .fs_flags = FS_REQUIRES_DEV, | |
107 | }; | |
108 | #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) | |
109 | #else | |
110 | #define IS_EXT3_SB(sb) (0) | |
111 | #endif | |
bd81d8ee | 112 | |
8fadc143 AR |
113 | ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, |
114 | struct ext4_group_desc *bg) | |
bd81d8ee | 115 | { |
3a14589c | 116 | return le32_to_cpu(bg->bg_block_bitmap_lo) | |
8fadc143 | 117 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 118 | (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); |
bd81d8ee LV |
119 | } |
120 | ||
8fadc143 AR |
121 | ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, |
122 | struct ext4_group_desc *bg) | |
bd81d8ee | 123 | { |
5272f837 | 124 | return le32_to_cpu(bg->bg_inode_bitmap_lo) | |
8fadc143 | 125 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 126 | (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); |
bd81d8ee LV |
127 | } |
128 | ||
8fadc143 AR |
129 | ext4_fsblk_t ext4_inode_table(struct super_block *sb, |
130 | struct ext4_group_desc *bg) | |
bd81d8ee | 131 | { |
5272f837 | 132 | return le32_to_cpu(bg->bg_inode_table_lo) | |
8fadc143 | 133 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 134 | (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); |
bd81d8ee LV |
135 | } |
136 | ||
560671a0 AK |
137 | __u32 ext4_free_blks_count(struct super_block *sb, |
138 | struct ext4_group_desc *bg) | |
139 | { | |
140 | return le16_to_cpu(bg->bg_free_blocks_count_lo) | | |
141 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 142 | (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); |
560671a0 AK |
143 | } |
144 | ||
145 | __u32 ext4_free_inodes_count(struct super_block *sb, | |
146 | struct ext4_group_desc *bg) | |
147 | { | |
148 | return le16_to_cpu(bg->bg_free_inodes_count_lo) | | |
149 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 150 | (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); |
560671a0 AK |
151 | } |
152 | ||
153 | __u32 ext4_used_dirs_count(struct super_block *sb, | |
154 | struct ext4_group_desc *bg) | |
155 | { | |
156 | return le16_to_cpu(bg->bg_used_dirs_count_lo) | | |
157 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 158 | (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); |
560671a0 AK |
159 | } |
160 | ||
161 | __u32 ext4_itable_unused_count(struct super_block *sb, | |
162 | struct ext4_group_desc *bg) | |
163 | { | |
164 | return le16_to_cpu(bg->bg_itable_unused_lo) | | |
165 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 166 | (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); |
560671a0 AK |
167 | } |
168 | ||
8fadc143 AR |
169 | void ext4_block_bitmap_set(struct super_block *sb, |
170 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 171 | { |
3a14589c | 172 | bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
173 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
174 | bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
175 | } |
176 | ||
8fadc143 AR |
177 | void ext4_inode_bitmap_set(struct super_block *sb, |
178 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 179 | { |
5272f837 | 180 | bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
181 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
182 | bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
183 | } |
184 | ||
8fadc143 AR |
185 | void ext4_inode_table_set(struct super_block *sb, |
186 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 187 | { |
5272f837 | 188 | bg->bg_inode_table_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
189 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
190 | bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
191 | } |
192 | ||
560671a0 AK |
193 | void ext4_free_blks_set(struct super_block *sb, |
194 | struct ext4_group_desc *bg, __u32 count) | |
195 | { | |
196 | bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); | |
197 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
198 | bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); | |
199 | } | |
200 | ||
201 | void ext4_free_inodes_set(struct super_block *sb, | |
202 | struct ext4_group_desc *bg, __u32 count) | |
203 | { | |
204 | bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); | |
205 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
206 | bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); | |
207 | } | |
208 | ||
209 | void ext4_used_dirs_set(struct super_block *sb, | |
210 | struct ext4_group_desc *bg, __u32 count) | |
211 | { | |
212 | bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); | |
213 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
214 | bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); | |
215 | } | |
216 | ||
217 | void ext4_itable_unused_set(struct super_block *sb, | |
218 | struct ext4_group_desc *bg, __u32 count) | |
219 | { | |
220 | bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); | |
221 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
222 | bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); | |
223 | } | |
224 | ||
d3d1faf6 CW |
225 | |
226 | /* Just increment the non-pointer handle value */ | |
227 | static handle_t *ext4_get_nojournal(void) | |
228 | { | |
229 | handle_t *handle = current->journal_info; | |
230 | unsigned long ref_cnt = (unsigned long)handle; | |
231 | ||
232 | BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT); | |
233 | ||
234 | ref_cnt++; | |
235 | handle = (handle_t *)ref_cnt; | |
236 | ||
237 | current->journal_info = handle; | |
238 | return handle; | |
239 | } | |
240 | ||
241 | ||
242 | /* Decrement the non-pointer handle value */ | |
243 | static void ext4_put_nojournal(handle_t *handle) | |
244 | { | |
245 | unsigned long ref_cnt = (unsigned long)handle; | |
246 | ||
247 | BUG_ON(ref_cnt == 0); | |
248 | ||
249 | ref_cnt--; | |
250 | handle = (handle_t *)ref_cnt; | |
251 | ||
252 | current->journal_info = handle; | |
253 | } | |
254 | ||
ac27a0ec | 255 | /* |
dab291af | 256 | * Wrappers for jbd2_journal_start/end. |
ac27a0ec DK |
257 | * |
258 | * The only special thing we need to do here is to make sure that all | |
259 | * journal_end calls result in the superblock being marked dirty, so | |
260 | * that sync() will call the filesystem's write_super callback if | |
261 | * appropriate. | |
be4f27d3 YY |
262 | * |
263 | * To avoid j_barrier hold in userspace when a user calls freeze(), | |
264 | * ext4 prevents a new handle from being started by s_frozen, which | |
265 | * is in an upper layer. | |
ac27a0ec | 266 | */ |
617ba13b | 267 | handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks) |
ac27a0ec DK |
268 | { |
269 | journal_t *journal; | |
be4f27d3 | 270 | handle_t *handle; |
ac27a0ec DK |
271 | |
272 | if (sb->s_flags & MS_RDONLY) | |
273 | return ERR_PTR(-EROFS); | |
274 | ||
617ba13b | 275 | journal = EXT4_SB(sb)->s_journal; |
be4f27d3 YY |
276 | handle = ext4_journal_current_handle(); |
277 | ||
278 | /* | |
279 | * If a handle has been started, it should be allowed to | |
280 | * finish, otherwise deadlock could happen between freeze | |
281 | * and others(e.g. truncate) due to the restart of the | |
282 | * journal handle if the filesystem is forzen and active | |
283 | * handles are not stopped. | |
284 | */ | |
285 | if (!handle) | |
286 | vfs_check_frozen(sb, SB_FREEZE_TRANS); | |
287 | ||
288 | if (!journal) | |
289 | return ext4_get_nojournal(); | |
290 | /* | |
291 | * Special case here: if the journal has aborted behind our | |
292 | * backs (eg. EIO in the commit thread), then we still need to | |
293 | * take the FS itself readonly cleanly. | |
294 | */ | |
295 | if (is_journal_aborted(journal)) { | |
296 | ext4_abort(sb, "Detected aborted journal"); | |
297 | return ERR_PTR(-EROFS); | |
ac27a0ec | 298 | } |
be4f27d3 | 299 | return jbd2_journal_start(journal, nblocks); |
ac27a0ec DK |
300 | } |
301 | ||
302 | /* | |
303 | * The only special thing we need to do here is to make sure that all | |
dab291af | 304 | * jbd2_journal_stop calls result in the superblock being marked dirty, so |
ac27a0ec DK |
305 | * that sync() will call the filesystem's write_super callback if |
306 | * appropriate. | |
307 | */ | |
c398eda0 | 308 | int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle) |
ac27a0ec DK |
309 | { |
310 | struct super_block *sb; | |
311 | int err; | |
312 | int rc; | |
313 | ||
0390131b | 314 | if (!ext4_handle_valid(handle)) { |
d3d1faf6 | 315 | ext4_put_nojournal(handle); |
0390131b FM |
316 | return 0; |
317 | } | |
ac27a0ec DK |
318 | sb = handle->h_transaction->t_journal->j_private; |
319 | err = handle->h_err; | |
dab291af | 320 | rc = jbd2_journal_stop(handle); |
ac27a0ec DK |
321 | |
322 | if (!err) | |
323 | err = rc; | |
324 | if (err) | |
c398eda0 | 325 | __ext4_std_error(sb, where, line, err); |
ac27a0ec DK |
326 | return err; |
327 | } | |
328 | ||
90c7201b TT |
329 | void ext4_journal_abort_handle(const char *caller, unsigned int line, |
330 | const char *err_fn, struct buffer_head *bh, | |
331 | handle_t *handle, int err) | |
ac27a0ec DK |
332 | { |
333 | char nbuf[16]; | |
617ba13b | 334 | const char *errstr = ext4_decode_error(NULL, err, nbuf); |
ac27a0ec | 335 | |
0390131b FM |
336 | BUG_ON(!ext4_handle_valid(handle)); |
337 | ||
ac27a0ec DK |
338 | if (bh) |
339 | BUFFER_TRACE(bh, "abort"); | |
340 | ||
341 | if (!handle->h_err) | |
342 | handle->h_err = err; | |
343 | ||
344 | if (is_handle_aborted(handle)) | |
345 | return; | |
346 | ||
90c7201b TT |
347 | printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n", |
348 | caller, line, errstr, err_fn); | |
ac27a0ec | 349 | |
dab291af | 350 | jbd2_journal_abort_handle(handle); |
ac27a0ec DK |
351 | } |
352 | ||
1c13d5c0 TT |
353 | static void __save_error_info(struct super_block *sb, const char *func, |
354 | unsigned int line) | |
355 | { | |
356 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; | |
357 | ||
358 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; | |
359 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
360 | es->s_last_error_time = cpu_to_le32(get_seconds()); | |
361 | strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func)); | |
362 | es->s_last_error_line = cpu_to_le32(line); | |
363 | if (!es->s_first_error_time) { | |
364 | es->s_first_error_time = es->s_last_error_time; | |
365 | strncpy(es->s_first_error_func, func, | |
366 | sizeof(es->s_first_error_func)); | |
367 | es->s_first_error_line = cpu_to_le32(line); | |
368 | es->s_first_error_ino = es->s_last_error_ino; | |
369 | es->s_first_error_block = es->s_last_error_block; | |
370 | } | |
66e61a9e TT |
371 | /* |
372 | * Start the daily error reporting function if it hasn't been | |
373 | * started already | |
374 | */ | |
375 | if (!es->s_error_count) | |
376 | mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ); | |
1c13d5c0 TT |
377 | es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1); |
378 | } | |
379 | ||
380 | static void save_error_info(struct super_block *sb, const char *func, | |
381 | unsigned int line) | |
382 | { | |
383 | __save_error_info(sb, func, line); | |
384 | ext4_commit_super(sb, 1); | |
385 | } | |
386 | ||
387 | ||
ac27a0ec DK |
388 | /* Deal with the reporting of failure conditions on a filesystem such as |
389 | * inconsistencies detected or read IO failures. | |
390 | * | |
391 | * On ext2, we can store the error state of the filesystem in the | |
617ba13b | 392 | * superblock. That is not possible on ext4, because we may have other |
ac27a0ec DK |
393 | * write ordering constraints on the superblock which prevent us from |
394 | * writing it out straight away; and given that the journal is about to | |
395 | * be aborted, we can't rely on the current, or future, transactions to | |
396 | * write out the superblock safely. | |
397 | * | |
dab291af | 398 | * We'll just use the jbd2_journal_abort() error code to record an error in |
d6b198bc | 399 | * the journal instead. On recovery, the journal will complain about |
ac27a0ec DK |
400 | * that error until we've noted it down and cleared it. |
401 | */ | |
402 | ||
617ba13b | 403 | static void ext4_handle_error(struct super_block *sb) |
ac27a0ec | 404 | { |
ac27a0ec DK |
405 | if (sb->s_flags & MS_RDONLY) |
406 | return; | |
407 | ||
2b2d6d01 | 408 | if (!test_opt(sb, ERRORS_CONT)) { |
617ba13b | 409 | journal_t *journal = EXT4_SB(sb)->s_journal; |
ac27a0ec | 410 | |
4ab2f15b | 411 | EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; |
ac27a0ec | 412 | if (journal) |
dab291af | 413 | jbd2_journal_abort(journal, -EIO); |
ac27a0ec | 414 | } |
2b2d6d01 | 415 | if (test_opt(sb, ERRORS_RO)) { |
b31e1552 | 416 | ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); |
ac27a0ec DK |
417 | sb->s_flags |= MS_RDONLY; |
418 | } | |
ac27a0ec | 419 | if (test_opt(sb, ERRORS_PANIC)) |
617ba13b | 420 | panic("EXT4-fs (device %s): panic forced after error\n", |
ac27a0ec DK |
421 | sb->s_id); |
422 | } | |
423 | ||
12062ddd | 424 | void __ext4_error(struct super_block *sb, const char *function, |
c398eda0 | 425 | unsigned int line, const char *fmt, ...) |
ac27a0ec | 426 | { |
0ff2ea7d | 427 | struct va_format vaf; |
ac27a0ec DK |
428 | va_list args; |
429 | ||
430 | va_start(args, fmt); | |
0ff2ea7d JP |
431 | vaf.fmt = fmt; |
432 | vaf.va = &args; | |
433 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", | |
434 | sb->s_id, function, line, current->comm, &vaf); | |
ac27a0ec DK |
435 | va_end(args); |
436 | ||
617ba13b | 437 | ext4_handle_error(sb); |
ac27a0ec DK |
438 | } |
439 | ||
c398eda0 TT |
440 | void ext4_error_inode(struct inode *inode, const char *function, |
441 | unsigned int line, ext4_fsblk_t block, | |
273df556 FM |
442 | const char *fmt, ...) |
443 | { | |
444 | va_list args; | |
f7c21177 | 445 | struct va_format vaf; |
1c13d5c0 | 446 | struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es; |
273df556 | 447 | |
1c13d5c0 TT |
448 | es->s_last_error_ino = cpu_to_le32(inode->i_ino); |
449 | es->s_last_error_block = cpu_to_le64(block); | |
450 | save_error_info(inode->i_sb, function, line); | |
273df556 | 451 | va_start(args, fmt); |
f7c21177 TT |
452 | vaf.fmt = fmt; |
453 | vaf.va = &args; | |
c398eda0 TT |
454 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ", |
455 | inode->i_sb->s_id, function, line, inode->i_ino); | |
456 | if (block) | |
f7c21177 TT |
457 | printk(KERN_CONT "block %llu: ", block); |
458 | printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf); | |
273df556 FM |
459 | va_end(args); |
460 | ||
461 | ext4_handle_error(inode->i_sb); | |
462 | } | |
463 | ||
c398eda0 | 464 | void ext4_error_file(struct file *file, const char *function, |
f7c21177 TT |
465 | unsigned int line, ext4_fsblk_t block, |
466 | const char *fmt, ...) | |
273df556 FM |
467 | { |
468 | va_list args; | |
f7c21177 | 469 | struct va_format vaf; |
1c13d5c0 | 470 | struct ext4_super_block *es; |
273df556 FM |
471 | struct inode *inode = file->f_dentry->d_inode; |
472 | char pathname[80], *path; | |
473 | ||
1c13d5c0 TT |
474 | es = EXT4_SB(inode->i_sb)->s_es; |
475 | es->s_last_error_ino = cpu_to_le32(inode->i_ino); | |
476 | save_error_info(inode->i_sb, function, line); | |
273df556 | 477 | path = d_path(&(file->f_path), pathname, sizeof(pathname)); |
f9a62d09 | 478 | if (IS_ERR(path)) |
273df556 FM |
479 | path = "(unknown)"; |
480 | printk(KERN_CRIT | |
f7c21177 TT |
481 | "EXT4-fs error (device %s): %s:%d: inode #%lu: ", |
482 | inode->i_sb->s_id, function, line, inode->i_ino); | |
483 | if (block) | |
484 | printk(KERN_CONT "block %llu: ", block); | |
485 | va_start(args, fmt); | |
486 | vaf.fmt = fmt; | |
487 | vaf.va = &args; | |
488 | printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf); | |
273df556 FM |
489 | va_end(args); |
490 | ||
491 | ext4_handle_error(inode->i_sb); | |
492 | } | |
493 | ||
2b2d6d01 | 494 | static const char *ext4_decode_error(struct super_block *sb, int errno, |
ac27a0ec DK |
495 | char nbuf[16]) |
496 | { | |
497 | char *errstr = NULL; | |
498 | ||
499 | switch (errno) { | |
500 | case -EIO: | |
501 | errstr = "IO failure"; | |
502 | break; | |
503 | case -ENOMEM: | |
504 | errstr = "Out of memory"; | |
505 | break; | |
506 | case -EROFS: | |
78f1ddbb TT |
507 | if (!sb || (EXT4_SB(sb)->s_journal && |
508 | EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) | |
ac27a0ec DK |
509 | errstr = "Journal has aborted"; |
510 | else | |
511 | errstr = "Readonly filesystem"; | |
512 | break; | |
513 | default: | |
514 | /* If the caller passed in an extra buffer for unknown | |
515 | * errors, textualise them now. Else we just return | |
516 | * NULL. */ | |
517 | if (nbuf) { | |
518 | /* Check for truncated error codes... */ | |
519 | if (snprintf(nbuf, 16, "error %d", -errno) >= 0) | |
520 | errstr = nbuf; | |
521 | } | |
522 | break; | |
523 | } | |
524 | ||
525 | return errstr; | |
526 | } | |
527 | ||
617ba13b | 528 | /* __ext4_std_error decodes expected errors from journaling functions |
ac27a0ec DK |
529 | * automatically and invokes the appropriate error response. */ |
530 | ||
c398eda0 TT |
531 | void __ext4_std_error(struct super_block *sb, const char *function, |
532 | unsigned int line, int errno) | |
ac27a0ec DK |
533 | { |
534 | char nbuf[16]; | |
535 | const char *errstr; | |
536 | ||
537 | /* Special case: if the error is EROFS, and we're not already | |
538 | * inside a transaction, then there's really no point in logging | |
539 | * an error. */ | |
540 | if (errno == -EROFS && journal_current_handle() == NULL && | |
541 | (sb->s_flags & MS_RDONLY)) | |
542 | return; | |
543 | ||
617ba13b | 544 | errstr = ext4_decode_error(sb, errno, nbuf); |
c398eda0 TT |
545 | printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", |
546 | sb->s_id, function, line, errstr); | |
1c13d5c0 | 547 | save_error_info(sb, function, line); |
ac27a0ec | 548 | |
617ba13b | 549 | ext4_handle_error(sb); |
ac27a0ec DK |
550 | } |
551 | ||
552 | /* | |
617ba13b | 553 | * ext4_abort is a much stronger failure handler than ext4_error. The |
ac27a0ec DK |
554 | * abort function may be used to deal with unrecoverable failures such |
555 | * as journal IO errors or ENOMEM at a critical moment in log management. | |
556 | * | |
557 | * We unconditionally force the filesystem into an ABORT|READONLY state, | |
558 | * unless the error response on the fs has been set to panic in which | |
559 | * case we take the easy way out and panic immediately. | |
560 | */ | |
561 | ||
c67d859e | 562 | void __ext4_abort(struct super_block *sb, const char *function, |
c398eda0 | 563 | unsigned int line, const char *fmt, ...) |
ac27a0ec DK |
564 | { |
565 | va_list args; | |
566 | ||
1c13d5c0 | 567 | save_error_info(sb, function, line); |
ac27a0ec | 568 | va_start(args, fmt); |
c398eda0 TT |
569 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id, |
570 | function, line); | |
ac27a0ec DK |
571 | vprintk(fmt, args); |
572 | printk("\n"); | |
573 | va_end(args); | |
574 | ||
1c13d5c0 TT |
575 | if ((sb->s_flags & MS_RDONLY) == 0) { |
576 | ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); | |
577 | sb->s_flags |= MS_RDONLY; | |
578 | EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; | |
579 | if (EXT4_SB(sb)->s_journal) | |
580 | jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); | |
581 | save_error_info(sb, function, line); | |
582 | } | |
ac27a0ec | 583 | if (test_opt(sb, ERRORS_PANIC)) |
617ba13b | 584 | panic("EXT4-fs panic from previous error\n"); |
ac27a0ec DK |
585 | } |
586 | ||
0ff2ea7d | 587 | void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...) |
b31e1552 | 588 | { |
0ff2ea7d | 589 | struct va_format vaf; |
b31e1552 ES |
590 | va_list args; |
591 | ||
592 | va_start(args, fmt); | |
0ff2ea7d JP |
593 | vaf.fmt = fmt; |
594 | vaf.va = &args; | |
595 | printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); | |
b31e1552 ES |
596 | va_end(args); |
597 | } | |
598 | ||
12062ddd | 599 | void __ext4_warning(struct super_block *sb, const char *function, |
c398eda0 | 600 | unsigned int line, const char *fmt, ...) |
ac27a0ec | 601 | { |
0ff2ea7d | 602 | struct va_format vaf; |
ac27a0ec DK |
603 | va_list args; |
604 | ||
605 | va_start(args, fmt); | |
0ff2ea7d JP |
606 | vaf.fmt = fmt; |
607 | vaf.va = &args; | |
608 | printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", | |
609 | sb->s_id, function, line, &vaf); | |
ac27a0ec DK |
610 | va_end(args); |
611 | } | |
612 | ||
e29136f8 TT |
613 | void __ext4_grp_locked_error(const char *function, unsigned int line, |
614 | struct super_block *sb, ext4_group_t grp, | |
615 | unsigned long ino, ext4_fsblk_t block, | |
616 | const char *fmt, ...) | |
5d1b1b3f AK |
617 | __releases(bitlock) |
618 | __acquires(bitlock) | |
619 | { | |
0ff2ea7d | 620 | struct va_format vaf; |
5d1b1b3f AK |
621 | va_list args; |
622 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; | |
623 | ||
1c13d5c0 TT |
624 | es->s_last_error_ino = cpu_to_le32(ino); |
625 | es->s_last_error_block = cpu_to_le64(block); | |
626 | __save_error_info(sb, function, line); | |
0ff2ea7d | 627 | |
5d1b1b3f | 628 | va_start(args, fmt); |
0ff2ea7d JP |
629 | |
630 | vaf.fmt = fmt; | |
631 | vaf.va = &args; | |
21149d61 | 632 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", |
e29136f8 TT |
633 | sb->s_id, function, line, grp); |
634 | if (ino) | |
0ff2ea7d | 635 | printk(KERN_CONT "inode %lu: ", ino); |
e29136f8 | 636 | if (block) |
0ff2ea7d JP |
637 | printk(KERN_CONT "block %llu:", (unsigned long long) block); |
638 | printk(KERN_CONT "%pV\n", &vaf); | |
5d1b1b3f AK |
639 | va_end(args); |
640 | ||
641 | if (test_opt(sb, ERRORS_CONT)) { | |
e2d67052 | 642 | ext4_commit_super(sb, 0); |
5d1b1b3f AK |
643 | return; |
644 | } | |
1c13d5c0 | 645 | |
5d1b1b3f AK |
646 | ext4_unlock_group(sb, grp); |
647 | ext4_handle_error(sb); | |
648 | /* | |
649 | * We only get here in the ERRORS_RO case; relocking the group | |
650 | * may be dangerous, but nothing bad will happen since the | |
651 | * filesystem will have already been marked read/only and the | |
652 | * journal has been aborted. We return 1 as a hint to callers | |
653 | * who might what to use the return value from | |
25985edc | 654 | * ext4_grp_locked_error() to distinguish between the |
5d1b1b3f AK |
655 | * ERRORS_CONT and ERRORS_RO case, and perhaps return more |
656 | * aggressively from the ext4 function in question, with a | |
657 | * more appropriate error code. | |
658 | */ | |
659 | ext4_lock_group(sb, grp); | |
660 | return; | |
661 | } | |
662 | ||
617ba13b | 663 | void ext4_update_dynamic_rev(struct super_block *sb) |
ac27a0ec | 664 | { |
617ba13b | 665 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
ac27a0ec | 666 | |
617ba13b | 667 | if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) |
ac27a0ec DK |
668 | return; |
669 | ||
12062ddd | 670 | ext4_warning(sb, |
ac27a0ec DK |
671 | "updating to rev %d because of new feature flag, " |
672 | "running e2fsck is recommended", | |
617ba13b | 673 | EXT4_DYNAMIC_REV); |
ac27a0ec | 674 | |
617ba13b MC |
675 | es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); |
676 | es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); | |
677 | es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); | |
ac27a0ec DK |
678 | /* leave es->s_feature_*compat flags alone */ |
679 | /* es->s_uuid will be set by e2fsck if empty */ | |
680 | ||
681 | /* | |
682 | * The rest of the superblock fields should be zero, and if not it | |
683 | * means they are likely already in use, so leave them alone. We | |
684 | * can leave it up to e2fsck to clean up any inconsistencies there. | |
685 | */ | |
686 | } | |
687 | ||
688 | /* | |
689 | * Open the external journal device | |
690 | */ | |
b31e1552 | 691 | static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) |
ac27a0ec DK |
692 | { |
693 | struct block_device *bdev; | |
694 | char b[BDEVNAME_SIZE]; | |
695 | ||
d4d77629 | 696 | bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); |
ac27a0ec DK |
697 | if (IS_ERR(bdev)) |
698 | goto fail; | |
699 | return bdev; | |
700 | ||
701 | fail: | |
b31e1552 | 702 | ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld", |
ac27a0ec DK |
703 | __bdevname(dev, b), PTR_ERR(bdev)); |
704 | return NULL; | |
705 | } | |
706 | ||
707 | /* | |
708 | * Release the journal device | |
709 | */ | |
617ba13b | 710 | static int ext4_blkdev_put(struct block_device *bdev) |
ac27a0ec | 711 | { |
e525fd89 | 712 | return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); |
ac27a0ec DK |
713 | } |
714 | ||
617ba13b | 715 | static int ext4_blkdev_remove(struct ext4_sb_info *sbi) |
ac27a0ec DK |
716 | { |
717 | struct block_device *bdev; | |
718 | int ret = -ENODEV; | |
719 | ||
720 | bdev = sbi->journal_bdev; | |
721 | if (bdev) { | |
617ba13b | 722 | ret = ext4_blkdev_put(bdev); |
ac27a0ec DK |
723 | sbi->journal_bdev = NULL; |
724 | } | |
725 | return ret; | |
726 | } | |
727 | ||
728 | static inline struct inode *orphan_list_entry(struct list_head *l) | |
729 | { | |
617ba13b | 730 | return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; |
ac27a0ec DK |
731 | } |
732 | ||
617ba13b | 733 | static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) |
ac27a0ec DK |
734 | { |
735 | struct list_head *l; | |
736 | ||
b31e1552 ES |
737 | ext4_msg(sb, KERN_ERR, "sb orphan head is %d", |
738 | le32_to_cpu(sbi->s_es->s_last_orphan)); | |
ac27a0ec DK |
739 | |
740 | printk(KERN_ERR "sb_info orphan list:\n"); | |
741 | list_for_each(l, &sbi->s_orphan) { | |
742 | struct inode *inode = orphan_list_entry(l); | |
743 | printk(KERN_ERR " " | |
744 | "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", | |
745 | inode->i_sb->s_id, inode->i_ino, inode, | |
746 | inode->i_mode, inode->i_nlink, | |
747 | NEXT_ORPHAN(inode)); | |
748 | } | |
749 | } | |
750 | ||
2b2d6d01 | 751 | static void ext4_put_super(struct super_block *sb) |
ac27a0ec | 752 | { |
617ba13b MC |
753 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
754 | struct ext4_super_block *es = sbi->s_es; | |
ef2cabf7 | 755 | int i, err; |
ac27a0ec | 756 | |
857ac889 | 757 | ext4_unregister_li_request(sb); |
e0ccfd95 CH |
758 | dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED); |
759 | ||
4c0425ff MC |
760 | flush_workqueue(sbi->dio_unwritten_wq); |
761 | destroy_workqueue(sbi->dio_unwritten_wq); | |
762 | ||
a9e220f8 | 763 | lock_super(sb); |
8c85e125 | 764 | if (sb->s_dirt) |
ebc1ac16 | 765 | ext4_commit_super(sb, 1); |
8c85e125 | 766 | |
0390131b FM |
767 | if (sbi->s_journal) { |
768 | err = jbd2_journal_destroy(sbi->s_journal); | |
769 | sbi->s_journal = NULL; | |
770 | if (err < 0) | |
c67d859e | 771 | ext4_abort(sb, "Couldn't clean up the journal"); |
0390131b | 772 | } |
d4edac31 | 773 | |
a1c6c569 | 774 | del_timer(&sbi->s_err_report); |
d4edac31 JB |
775 | ext4_release_system_zone(sb); |
776 | ext4_mb_release(sb); | |
777 | ext4_ext_release(sb); | |
778 | ext4_xattr_put_super(sb); | |
779 | ||
ac27a0ec | 780 | if (!(sb->s_flags & MS_RDONLY)) { |
617ba13b | 781 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); |
ac27a0ec | 782 | es->s_state = cpu_to_le16(sbi->s_mount_state); |
e2d67052 | 783 | ext4_commit_super(sb, 1); |
ac27a0ec | 784 | } |
240799cd | 785 | if (sbi->s_proc) { |
9f6200bb | 786 | remove_proc_entry(sb->s_id, ext4_proc_root); |
240799cd | 787 | } |
3197ebdb | 788 | kobject_del(&sbi->s_kobj); |
ac27a0ec DK |
789 | |
790 | for (i = 0; i < sbi->s_gdb_count; i++) | |
791 | brelse(sbi->s_group_desc[i]); | |
792 | kfree(sbi->s_group_desc); | |
c5ca7c76 TT |
793 | if (is_vmalloc_addr(sbi->s_flex_groups)) |
794 | vfree(sbi->s_flex_groups); | |
795 | else | |
796 | kfree(sbi->s_flex_groups); | |
ac27a0ec DK |
797 | percpu_counter_destroy(&sbi->s_freeblocks_counter); |
798 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
799 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
6bc6e63f | 800 | percpu_counter_destroy(&sbi->s_dirtyblocks_counter); |
ac27a0ec DK |
801 | brelse(sbi->s_sbh); |
802 | #ifdef CONFIG_QUOTA | |
803 | for (i = 0; i < MAXQUOTAS; i++) | |
804 | kfree(sbi->s_qf_names[i]); | |
805 | #endif | |
806 | ||
807 | /* Debugging code just in case the in-memory inode orphan list | |
808 | * isn't empty. The on-disk one can be non-empty if we've | |
809 | * detected an error and taken the fs readonly, but the | |
810 | * in-memory list had better be clean by this point. */ | |
811 | if (!list_empty(&sbi->s_orphan)) | |
812 | dump_orphan_list(sb, sbi); | |
813 | J_ASSERT(list_empty(&sbi->s_orphan)); | |
814 | ||
f98393a6 | 815 | invalidate_bdev(sb->s_bdev); |
ac27a0ec DK |
816 | if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { |
817 | /* | |
818 | * Invalidate the journal device's buffers. We don't want them | |
819 | * floating about in memory - the physical journal device may | |
820 | * hotswapped, and it breaks the `ro-after' testing code. | |
821 | */ | |
822 | sync_blockdev(sbi->journal_bdev); | |
f98393a6 | 823 | invalidate_bdev(sbi->journal_bdev); |
617ba13b | 824 | ext4_blkdev_remove(sbi); |
ac27a0ec | 825 | } |
c5e06d10 JL |
826 | if (sbi->s_mmp_tsk) |
827 | kthread_stop(sbi->s_mmp_tsk); | |
ac27a0ec | 828 | sb->s_fs_info = NULL; |
3197ebdb TT |
829 | /* |
830 | * Now that we are completely done shutting down the | |
831 | * superblock, we need to actually destroy the kobject. | |
832 | */ | |
3197ebdb TT |
833 | unlock_super(sb); |
834 | kobject_put(&sbi->s_kobj); | |
835 | wait_for_completion(&sbi->s_kobj_unregister); | |
705895b6 | 836 | kfree(sbi->s_blockgroup_lock); |
ac27a0ec | 837 | kfree(sbi); |
ac27a0ec DK |
838 | } |
839 | ||
e18b890b | 840 | static struct kmem_cache *ext4_inode_cachep; |
ac27a0ec DK |
841 | |
842 | /* | |
843 | * Called inside transaction, so use GFP_NOFS | |
844 | */ | |
617ba13b | 845 | static struct inode *ext4_alloc_inode(struct super_block *sb) |
ac27a0ec | 846 | { |
617ba13b | 847 | struct ext4_inode_info *ei; |
ac27a0ec | 848 | |
e6b4f8da | 849 | ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); |
ac27a0ec DK |
850 | if (!ei) |
851 | return NULL; | |
0b8e58a1 | 852 | |
ac27a0ec | 853 | ei->vfs_inode.i_version = 1; |
91246c00 | 854 | ei->vfs_inode.i_data.writeback_index = 0; |
a86c6181 | 855 | memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache)); |
c9de560d AT |
856 | INIT_LIST_HEAD(&ei->i_prealloc_list); |
857 | spin_lock_init(&ei->i_prealloc_lock); | |
d2a17637 MC |
858 | ei->i_reserved_data_blocks = 0; |
859 | ei->i_reserved_meta_blocks = 0; | |
860 | ei->i_allocated_meta_blocks = 0; | |
9d0be502 | 861 | ei->i_da_metadata_calc_len = 0; |
d2a17637 | 862 | spin_lock_init(&(ei->i_block_reservation_lock)); |
a9e7f447 DM |
863 | #ifdef CONFIG_QUOTA |
864 | ei->i_reserved_quota = 0; | |
865 | #endif | |
8aefcd55 | 866 | ei->jinode = NULL; |
c7064ef1 | 867 | INIT_LIST_HEAD(&ei->i_completed_io_list); |
744692dc | 868 | spin_lock_init(&ei->i_completed_io_lock); |
8d5d02e6 | 869 | ei->cur_aio_dio = NULL; |
b436b9be JK |
870 | ei->i_sync_tid = 0; |
871 | ei->i_datasync_tid = 0; | |
f7ad6d2e | 872 | atomic_set(&ei->i_ioend_count, 0); |
e9e3bcec | 873 | atomic_set(&ei->i_aiodio_unwritten, 0); |
0b8e58a1 | 874 | |
ac27a0ec DK |
875 | return &ei->vfs_inode; |
876 | } | |
877 | ||
7ff9c073 TT |
878 | static int ext4_drop_inode(struct inode *inode) |
879 | { | |
880 | int drop = generic_drop_inode(inode); | |
881 | ||
882 | trace_ext4_drop_inode(inode, drop); | |
883 | return drop; | |
884 | } | |
885 | ||
fa0d7e3d NP |
886 | static void ext4_i_callback(struct rcu_head *head) |
887 | { | |
888 | struct inode *inode = container_of(head, struct inode, i_rcu); | |
889 | INIT_LIST_HEAD(&inode->i_dentry); | |
890 | kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); | |
891 | } | |
892 | ||
617ba13b | 893 | static void ext4_destroy_inode(struct inode *inode) |
ac27a0ec | 894 | { |
f7ad6d2e | 895 | ext4_ioend_wait(inode); |
9f7dd93d | 896 | if (!list_empty(&(EXT4_I(inode)->i_orphan))) { |
b31e1552 ES |
897 | ext4_msg(inode->i_sb, KERN_ERR, |
898 | "Inode %lu (%p): orphan list check failed!", | |
899 | inode->i_ino, EXT4_I(inode)); | |
9f7dd93d VA |
900 | print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, |
901 | EXT4_I(inode), sizeof(struct ext4_inode_info), | |
902 | true); | |
903 | dump_stack(); | |
904 | } | |
fa0d7e3d | 905 | call_rcu(&inode->i_rcu, ext4_i_callback); |
ac27a0ec DK |
906 | } |
907 | ||
51cc5068 | 908 | static void init_once(void *foo) |
ac27a0ec | 909 | { |
617ba13b | 910 | struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; |
ac27a0ec | 911 | |
a35afb83 | 912 | INIT_LIST_HEAD(&ei->i_orphan); |
03010a33 | 913 | #ifdef CONFIG_EXT4_FS_XATTR |
a35afb83 | 914 | init_rwsem(&ei->xattr_sem); |
ac27a0ec | 915 | #endif |
0e855ac8 | 916 | init_rwsem(&ei->i_data_sem); |
a35afb83 | 917 | inode_init_once(&ei->vfs_inode); |
ac27a0ec DK |
918 | } |
919 | ||
920 | static int init_inodecache(void) | |
921 | { | |
617ba13b MC |
922 | ext4_inode_cachep = kmem_cache_create("ext4_inode_cache", |
923 | sizeof(struct ext4_inode_info), | |
ac27a0ec DK |
924 | 0, (SLAB_RECLAIM_ACCOUNT| |
925 | SLAB_MEM_SPREAD), | |
20c2df83 | 926 | init_once); |
617ba13b | 927 | if (ext4_inode_cachep == NULL) |
ac27a0ec DK |
928 | return -ENOMEM; |
929 | return 0; | |
930 | } | |
931 | ||
932 | static void destroy_inodecache(void) | |
933 | { | |
617ba13b | 934 | kmem_cache_destroy(ext4_inode_cachep); |
ac27a0ec DK |
935 | } |
936 | ||
0930fcc1 | 937 | void ext4_clear_inode(struct inode *inode) |
ac27a0ec | 938 | { |
0930fcc1 AV |
939 | invalidate_inode_buffers(inode); |
940 | end_writeback(inode); | |
9f754758 | 941 | dquot_drop(inode); |
c2ea3fde | 942 | ext4_discard_preallocations(inode); |
8aefcd55 TT |
943 | if (EXT4_I(inode)->jinode) { |
944 | jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), | |
945 | EXT4_I(inode)->jinode); | |
946 | jbd2_free_inode(EXT4_I(inode)->jinode); | |
947 | EXT4_I(inode)->jinode = NULL; | |
948 | } | |
ac27a0ec DK |
949 | } |
950 | ||
2b2d6d01 TT |
951 | static inline void ext4_show_quota_options(struct seq_file *seq, |
952 | struct super_block *sb) | |
ac27a0ec DK |
953 | { |
954 | #if defined(CONFIG_QUOTA) | |
617ba13b | 955 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 956 | |
5a20bdfc JK |
957 | if (sbi->s_jquota_fmt) { |
958 | char *fmtname = ""; | |
959 | ||
960 | switch (sbi->s_jquota_fmt) { | |
961 | case QFMT_VFS_OLD: | |
962 | fmtname = "vfsold"; | |
963 | break; | |
964 | case QFMT_VFS_V0: | |
965 | fmtname = "vfsv0"; | |
966 | break; | |
967 | case QFMT_VFS_V1: | |
968 | fmtname = "vfsv1"; | |
969 | break; | |
970 | } | |
971 | seq_printf(seq, ",jqfmt=%s", fmtname); | |
972 | } | |
ac27a0ec DK |
973 | |
974 | if (sbi->s_qf_names[USRQUOTA]) | |
975 | seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]); | |
976 | ||
977 | if (sbi->s_qf_names[GRPQUOTA]) | |
978 | seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]); | |
979 | ||
482a7425 | 980 | if (test_opt(sb, USRQUOTA)) |
ac27a0ec DK |
981 | seq_puts(seq, ",usrquota"); |
982 | ||
482a7425 | 983 | if (test_opt(sb, GRPQUOTA)) |
ac27a0ec DK |
984 | seq_puts(seq, ",grpquota"); |
985 | #endif | |
986 | } | |
987 | ||
d9c9bef1 MS |
988 | /* |
989 | * Show an option if | |
990 | * - it's set to a non-default value OR | |
991 | * - if the per-sb default is different from the global default | |
992 | */ | |
617ba13b | 993 | static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs) |
ac27a0ec | 994 | { |
aa22df2c AK |
995 | int def_errors; |
996 | unsigned long def_mount_opts; | |
ac27a0ec | 997 | struct super_block *sb = vfs->mnt_sb; |
d9c9bef1 MS |
998 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
999 | struct ext4_super_block *es = sbi->s_es; | |
d9c9bef1 MS |
1000 | |
1001 | def_mount_opts = le32_to_cpu(es->s_default_mount_opts); | |
aa22df2c | 1002 | def_errors = le16_to_cpu(es->s_errors); |
d9c9bef1 MS |
1003 | |
1004 | if (sbi->s_sb_block != 1) | |
1005 | seq_printf(seq, ",sb=%llu", sbi->s_sb_block); | |
1006 | if (test_opt(sb, MINIX_DF)) | |
1007 | seq_puts(seq, ",minixdf"); | |
aa22df2c | 1008 | if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS)) |
d9c9bef1 MS |
1009 | seq_puts(seq, ",grpid"); |
1010 | if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS)) | |
1011 | seq_puts(seq, ",nogrpid"); | |
1012 | if (sbi->s_resuid != EXT4_DEF_RESUID || | |
1013 | le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) { | |
1014 | seq_printf(seq, ",resuid=%u", sbi->s_resuid); | |
1015 | } | |
1016 | if (sbi->s_resgid != EXT4_DEF_RESGID || | |
1017 | le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) { | |
1018 | seq_printf(seq, ",resgid=%u", sbi->s_resgid); | |
1019 | } | |
bb4f397a | 1020 | if (test_opt(sb, ERRORS_RO)) { |
d9c9bef1 | 1021 | if (def_errors == EXT4_ERRORS_PANIC || |
bb4f397a AK |
1022 | def_errors == EXT4_ERRORS_CONTINUE) { |
1023 | seq_puts(seq, ",errors=remount-ro"); | |
d9c9bef1 MS |
1024 | } |
1025 | } | |
aa22df2c | 1026 | if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) |
bb4f397a | 1027 | seq_puts(seq, ",errors=continue"); |
aa22df2c | 1028 | if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) |
d9c9bef1 | 1029 | seq_puts(seq, ",errors=panic"); |
aa22df2c | 1030 | if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16)) |
d9c9bef1 | 1031 | seq_puts(seq, ",nouid32"); |
aa22df2c | 1032 | if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG)) |
d9c9bef1 MS |
1033 | seq_puts(seq, ",debug"); |
1034 | if (test_opt(sb, OLDALLOC)) | |
1035 | seq_puts(seq, ",oldalloc"); | |
03010a33 | 1036 | #ifdef CONFIG_EXT4_FS_XATTR |
ea663336 | 1037 | if (test_opt(sb, XATTR_USER)) |
d9c9bef1 | 1038 | seq_puts(seq, ",user_xattr"); |
ea663336 | 1039 | if (!test_opt(sb, XATTR_USER)) |
d9c9bef1 | 1040 | seq_puts(seq, ",nouser_xattr"); |
d9c9bef1 | 1041 | #endif |
03010a33 | 1042 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
aa22df2c | 1043 | if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL)) |
d9c9bef1 MS |
1044 | seq_puts(seq, ",acl"); |
1045 | if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL)) | |
1046 | seq_puts(seq, ",noacl"); | |
1047 | #endif | |
30773840 | 1048 | if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { |
d9c9bef1 MS |
1049 | seq_printf(seq, ",commit=%u", |
1050 | (unsigned) (sbi->s_commit_interval / HZ)); | |
1051 | } | |
30773840 TT |
1052 | if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) { |
1053 | seq_printf(seq, ",min_batch_time=%u", | |
1054 | (unsigned) sbi->s_min_batch_time); | |
1055 | } | |
1056 | if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) { | |
1057 | seq_printf(seq, ",max_batch_time=%u", | |
1058 | (unsigned) sbi->s_min_batch_time); | |
1059 | } | |
1060 | ||
571640ca ES |
1061 | /* |
1062 | * We're changing the default of barrier mount option, so | |
1063 | * let's always display its mount state so it's clear what its | |
1064 | * status is. | |
1065 | */ | |
1066 | seq_puts(seq, ",barrier="); | |
1067 | seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0"); | |
cd0b6a39 TT |
1068 | if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) |
1069 | seq_puts(seq, ",journal_async_commit"); | |
39a4bade JK |
1070 | else if (test_opt(sb, JOURNAL_CHECKSUM)) |
1071 | seq_puts(seq, ",journal_checksum"); | |
25ec56b5 JNC |
1072 | if (test_opt(sb, I_VERSION)) |
1073 | seq_puts(seq, ",i_version"); | |
8b67f04a TT |
1074 | if (!test_opt(sb, DELALLOC) && |
1075 | !(def_mount_opts & EXT4_DEFM_NODELALLOC)) | |
dd919b98 AK |
1076 | seq_puts(seq, ",nodelalloc"); |
1077 | ||
6fd7a467 TT |
1078 | if (!test_opt(sb, MBLK_IO_SUBMIT)) |
1079 | seq_puts(seq, ",nomblk_io_submit"); | |
cb45bbe4 MS |
1080 | if (sbi->s_stripe) |
1081 | seq_printf(seq, ",stripe=%lu", sbi->s_stripe); | |
aa22df2c AK |
1082 | /* |
1083 | * journal mode get enabled in different ways | |
1084 | * So just print the value even if we didn't specify it | |
1085 | */ | |
617ba13b | 1086 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) |
ac27a0ec | 1087 | seq_puts(seq, ",data=journal"); |
617ba13b | 1088 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) |
ac27a0ec | 1089 | seq_puts(seq, ",data=ordered"); |
617ba13b | 1090 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) |
ac27a0ec DK |
1091 | seq_puts(seq, ",data=writeback"); |
1092 | ||
240799cd TT |
1093 | if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) |
1094 | seq_printf(seq, ",inode_readahead_blks=%u", | |
1095 | sbi->s_inode_readahead_blks); | |
1096 | ||
5bf5683a HK |
1097 | if (test_opt(sb, DATA_ERR_ABORT)) |
1098 | seq_puts(seq, ",data_err=abort"); | |
1099 | ||
afd4672d | 1100 | if (test_opt(sb, NO_AUTO_DA_ALLOC)) |
06705bff | 1101 | seq_puts(seq, ",noauto_da_alloc"); |
afd4672d | 1102 | |
8b67f04a | 1103 | if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD)) |
5328e635 ES |
1104 | seq_puts(seq, ",discard"); |
1105 | ||
e3bb52ae ES |
1106 | if (test_opt(sb, NOLOAD)) |
1107 | seq_puts(seq, ",norecovery"); | |
1108 | ||
744692dc JZ |
1109 | if (test_opt(sb, DIOREAD_NOLOCK)) |
1110 | seq_puts(seq, ",dioread_nolock"); | |
1111 | ||
8b67f04a TT |
1112 | if (test_opt(sb, BLOCK_VALIDITY) && |
1113 | !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)) | |
1114 | seq_puts(seq, ",block_validity"); | |
1115 | ||
bfff6873 LC |
1116 | if (!test_opt(sb, INIT_INODE_TABLE)) |
1117 | seq_puts(seq, ",noinit_inode_table"); | |
51ce6511 | 1118 | else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT) |
bfff6873 LC |
1119 | seq_printf(seq, ",init_inode_table=%u", |
1120 | (unsigned) sbi->s_li_wait_mult); | |
1121 | ||
617ba13b | 1122 | ext4_show_quota_options(seq, sb); |
0b8e58a1 | 1123 | |
ac27a0ec DK |
1124 | return 0; |
1125 | } | |
1126 | ||
1b961ac0 | 1127 | static struct inode *ext4_nfs_get_inode(struct super_block *sb, |
0b8e58a1 | 1128 | u64 ino, u32 generation) |
ac27a0ec | 1129 | { |
ac27a0ec | 1130 | struct inode *inode; |
ac27a0ec | 1131 | |
617ba13b | 1132 | if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO) |
ac27a0ec | 1133 | return ERR_PTR(-ESTALE); |
617ba13b | 1134 | if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count)) |
ac27a0ec DK |
1135 | return ERR_PTR(-ESTALE); |
1136 | ||
1137 | /* iget isn't really right if the inode is currently unallocated!! | |
1138 | * | |
617ba13b | 1139 | * ext4_read_inode will return a bad_inode if the inode had been |
ac27a0ec DK |
1140 | * deleted, so we should be safe. |
1141 | * | |
1142 | * Currently we don't know the generation for parent directory, so | |
1143 | * a generation of 0 means "accept any" | |
1144 | */ | |
1d1fe1ee DH |
1145 | inode = ext4_iget(sb, ino); |
1146 | if (IS_ERR(inode)) | |
1147 | return ERR_CAST(inode); | |
1148 | if (generation && inode->i_generation != generation) { | |
ac27a0ec DK |
1149 | iput(inode); |
1150 | return ERR_PTR(-ESTALE); | |
1151 | } | |
1b961ac0 CH |
1152 | |
1153 | return inode; | |
1154 | } | |
1155 | ||
1156 | static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, | |
0b8e58a1 | 1157 | int fh_len, int fh_type) |
1b961ac0 CH |
1158 | { |
1159 | return generic_fh_to_dentry(sb, fid, fh_len, fh_type, | |
1160 | ext4_nfs_get_inode); | |
1161 | } | |
1162 | ||
1163 | static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, | |
0b8e58a1 | 1164 | int fh_len, int fh_type) |
1b961ac0 CH |
1165 | { |
1166 | return generic_fh_to_parent(sb, fid, fh_len, fh_type, | |
1167 | ext4_nfs_get_inode); | |
ac27a0ec DK |
1168 | } |
1169 | ||
c39a7f84 TO |
1170 | /* |
1171 | * Try to release metadata pages (indirect blocks, directories) which are | |
1172 | * mapped via the block device. Since these pages could have journal heads | |
1173 | * which would prevent try_to_free_buffers() from freeing them, we must use | |
1174 | * jbd2 layer's try_to_free_buffers() function to release them. | |
1175 | */ | |
0b8e58a1 AD |
1176 | static int bdev_try_to_free_page(struct super_block *sb, struct page *page, |
1177 | gfp_t wait) | |
c39a7f84 TO |
1178 | { |
1179 | journal_t *journal = EXT4_SB(sb)->s_journal; | |
1180 | ||
1181 | WARN_ON(PageChecked(page)); | |
1182 | if (!page_has_buffers(page)) | |
1183 | return 0; | |
1184 | if (journal) | |
1185 | return jbd2_journal_try_to_free_buffers(journal, page, | |
1186 | wait & ~__GFP_WAIT); | |
1187 | return try_to_free_buffers(page); | |
1188 | } | |
1189 | ||
ac27a0ec | 1190 | #ifdef CONFIG_QUOTA |
af5bc92d | 1191 | #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group") |
2b2d6d01 | 1192 | #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA)) |
ac27a0ec | 1193 | |
617ba13b MC |
1194 | static int ext4_write_dquot(struct dquot *dquot); |
1195 | static int ext4_acquire_dquot(struct dquot *dquot); | |
1196 | static int ext4_release_dquot(struct dquot *dquot); | |
1197 | static int ext4_mark_dquot_dirty(struct dquot *dquot); | |
1198 | static int ext4_write_info(struct super_block *sb, int type); | |
6f28e087 | 1199 | static int ext4_quota_on(struct super_block *sb, int type, int format_id, |
f00c9e44 | 1200 | struct path *path); |
ca0e05e4 | 1201 | static int ext4_quota_off(struct super_block *sb, int type); |
617ba13b MC |
1202 | static int ext4_quota_on_mount(struct super_block *sb, int type); |
1203 | static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, | |
ac27a0ec | 1204 | size_t len, loff_t off); |
617ba13b | 1205 | static ssize_t ext4_quota_write(struct super_block *sb, int type, |
ac27a0ec DK |
1206 | const char *data, size_t len, loff_t off); |
1207 | ||
61e225dc | 1208 | static const struct dquot_operations ext4_quota_operations = { |
60e58e0f | 1209 | .get_reserved_space = ext4_get_reserved_space, |
617ba13b MC |
1210 | .write_dquot = ext4_write_dquot, |
1211 | .acquire_dquot = ext4_acquire_dquot, | |
1212 | .release_dquot = ext4_release_dquot, | |
1213 | .mark_dirty = ext4_mark_dquot_dirty, | |
a5b5ee32 JK |
1214 | .write_info = ext4_write_info, |
1215 | .alloc_dquot = dquot_alloc, | |
1216 | .destroy_dquot = dquot_destroy, | |
ac27a0ec DK |
1217 | }; |
1218 | ||
0d54b217 | 1219 | static const struct quotactl_ops ext4_qctl_operations = { |
617ba13b | 1220 | .quota_on = ext4_quota_on, |
ca0e05e4 | 1221 | .quota_off = ext4_quota_off, |
287a8095 CH |
1222 | .quota_sync = dquot_quota_sync, |
1223 | .get_info = dquot_get_dqinfo, | |
1224 | .set_info = dquot_set_dqinfo, | |
1225 | .get_dqblk = dquot_get_dqblk, | |
1226 | .set_dqblk = dquot_set_dqblk | |
ac27a0ec DK |
1227 | }; |
1228 | #endif | |
1229 | ||
ee9b6d61 | 1230 | static const struct super_operations ext4_sops = { |
617ba13b MC |
1231 | .alloc_inode = ext4_alloc_inode, |
1232 | .destroy_inode = ext4_destroy_inode, | |
617ba13b MC |
1233 | .write_inode = ext4_write_inode, |
1234 | .dirty_inode = ext4_dirty_inode, | |
7ff9c073 | 1235 | .drop_inode = ext4_drop_inode, |
0930fcc1 | 1236 | .evict_inode = ext4_evict_inode, |
617ba13b | 1237 | .put_super = ext4_put_super, |
617ba13b | 1238 | .sync_fs = ext4_sync_fs, |
c4be0c1d TS |
1239 | .freeze_fs = ext4_freeze, |
1240 | .unfreeze_fs = ext4_unfreeze, | |
617ba13b MC |
1241 | .statfs = ext4_statfs, |
1242 | .remount_fs = ext4_remount, | |
617ba13b | 1243 | .show_options = ext4_show_options, |
ac27a0ec | 1244 | #ifdef CONFIG_QUOTA |
617ba13b MC |
1245 | .quota_read = ext4_quota_read, |
1246 | .quota_write = ext4_quota_write, | |
ac27a0ec | 1247 | #endif |
c39a7f84 | 1248 | .bdev_try_to_free_page = bdev_try_to_free_page, |
ac27a0ec DK |
1249 | }; |
1250 | ||
9ca92389 TT |
1251 | static const struct super_operations ext4_nojournal_sops = { |
1252 | .alloc_inode = ext4_alloc_inode, | |
1253 | .destroy_inode = ext4_destroy_inode, | |
1254 | .write_inode = ext4_write_inode, | |
1255 | .dirty_inode = ext4_dirty_inode, | |
7ff9c073 | 1256 | .drop_inode = ext4_drop_inode, |
0930fcc1 | 1257 | .evict_inode = ext4_evict_inode, |
9ca92389 TT |
1258 | .write_super = ext4_write_super, |
1259 | .put_super = ext4_put_super, | |
1260 | .statfs = ext4_statfs, | |
1261 | .remount_fs = ext4_remount, | |
9ca92389 TT |
1262 | .show_options = ext4_show_options, |
1263 | #ifdef CONFIG_QUOTA | |
1264 | .quota_read = ext4_quota_read, | |
1265 | .quota_write = ext4_quota_write, | |
1266 | #endif | |
1267 | .bdev_try_to_free_page = bdev_try_to_free_page, | |
1268 | }; | |
1269 | ||
39655164 | 1270 | static const struct export_operations ext4_export_ops = { |
1b961ac0 CH |
1271 | .fh_to_dentry = ext4_fh_to_dentry, |
1272 | .fh_to_parent = ext4_fh_to_parent, | |
617ba13b | 1273 | .get_parent = ext4_get_parent, |
ac27a0ec DK |
1274 | }; |
1275 | ||
1276 | enum { | |
1277 | Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, | |
1278 | Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, | |
01436ef2 | 1279 | Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov, |
ac27a0ec | 1280 | Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, |
06705bff | 1281 | Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh, |
30773840 | 1282 | Opt_commit, Opt_min_batch_time, Opt_max_batch_time, |
c3191067 | 1283 | Opt_journal_update, Opt_journal_dev, |
818d276c | 1284 | Opt_journal_checksum, Opt_journal_async_commit, |
ac27a0ec | 1285 | Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, |
296c355c | 1286 | Opt_data_err_abort, Opt_data_err_ignore, |
ac27a0ec | 1287 | Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, |
5a20bdfc JK |
1288 | Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, |
1289 | Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, | |
1290 | Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version, | |
1449032b TT |
1291 | Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit, |
1292 | Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, | |
5328e635 | 1293 | Opt_inode_readahead_blks, Opt_journal_ioprio, |
744692dc | 1294 | Opt_dioread_nolock, Opt_dioread_lock, |
5328e635 | 1295 | Opt_discard, Opt_nodiscard, |
bfff6873 | 1296 | Opt_init_inode_table, Opt_noinit_inode_table, |
ac27a0ec DK |
1297 | }; |
1298 | ||
a447c093 | 1299 | static const match_table_t tokens = { |
ac27a0ec DK |
1300 | {Opt_bsd_df, "bsddf"}, |
1301 | {Opt_minix_df, "minixdf"}, | |
1302 | {Opt_grpid, "grpid"}, | |
1303 | {Opt_grpid, "bsdgroups"}, | |
1304 | {Opt_nogrpid, "nogrpid"}, | |
1305 | {Opt_nogrpid, "sysvgroups"}, | |
1306 | {Opt_resgid, "resgid=%u"}, | |
1307 | {Opt_resuid, "resuid=%u"}, | |
1308 | {Opt_sb, "sb=%u"}, | |
1309 | {Opt_err_cont, "errors=continue"}, | |
1310 | {Opt_err_panic, "errors=panic"}, | |
1311 | {Opt_err_ro, "errors=remount-ro"}, | |
1312 | {Opt_nouid32, "nouid32"}, | |
ac27a0ec DK |
1313 | {Opt_debug, "debug"}, |
1314 | {Opt_oldalloc, "oldalloc"}, | |
1315 | {Opt_orlov, "orlov"}, | |
1316 | {Opt_user_xattr, "user_xattr"}, | |
1317 | {Opt_nouser_xattr, "nouser_xattr"}, | |
1318 | {Opt_acl, "acl"}, | |
1319 | {Opt_noacl, "noacl"}, | |
ac27a0ec | 1320 | {Opt_noload, "noload"}, |
e3bb52ae | 1321 | {Opt_noload, "norecovery"}, |
ac27a0ec DK |
1322 | {Opt_nobh, "nobh"}, |
1323 | {Opt_bh, "bh"}, | |
1324 | {Opt_commit, "commit=%u"}, | |
30773840 TT |
1325 | {Opt_min_batch_time, "min_batch_time=%u"}, |
1326 | {Opt_max_batch_time, "max_batch_time=%u"}, | |
ac27a0ec | 1327 | {Opt_journal_update, "journal=update"}, |
ac27a0ec | 1328 | {Opt_journal_dev, "journal_dev=%u"}, |
818d276c GS |
1329 | {Opt_journal_checksum, "journal_checksum"}, |
1330 | {Opt_journal_async_commit, "journal_async_commit"}, | |
ac27a0ec DK |
1331 | {Opt_abort, "abort"}, |
1332 | {Opt_data_journal, "data=journal"}, | |
1333 | {Opt_data_ordered, "data=ordered"}, | |
1334 | {Opt_data_writeback, "data=writeback"}, | |
5bf5683a HK |
1335 | {Opt_data_err_abort, "data_err=abort"}, |
1336 | {Opt_data_err_ignore, "data_err=ignore"}, | |
ac27a0ec DK |
1337 | {Opt_offusrjquota, "usrjquota="}, |
1338 | {Opt_usrjquota, "usrjquota=%s"}, | |
1339 | {Opt_offgrpjquota, "grpjquota="}, | |
1340 | {Opt_grpjquota, "grpjquota=%s"}, | |
1341 | {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, | |
1342 | {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, | |
5a20bdfc | 1343 | {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, |
ac27a0ec DK |
1344 | {Opt_grpquota, "grpquota"}, |
1345 | {Opt_noquota, "noquota"}, | |
1346 | {Opt_quota, "quota"}, | |
1347 | {Opt_usrquota, "usrquota"}, | |
1348 | {Opt_barrier, "barrier=%u"}, | |
06705bff TT |
1349 | {Opt_barrier, "barrier"}, |
1350 | {Opt_nobarrier, "nobarrier"}, | |
25ec56b5 | 1351 | {Opt_i_version, "i_version"}, |
c9de560d | 1352 | {Opt_stripe, "stripe=%u"}, |
ac27a0ec | 1353 | {Opt_resize, "resize"}, |
64769240 | 1354 | {Opt_delalloc, "delalloc"}, |
dd919b98 | 1355 | {Opt_nodelalloc, "nodelalloc"}, |
1449032b TT |
1356 | {Opt_mblk_io_submit, "mblk_io_submit"}, |
1357 | {Opt_nomblk_io_submit, "nomblk_io_submit"}, | |
6fd058f7 TT |
1358 | {Opt_block_validity, "block_validity"}, |
1359 | {Opt_noblock_validity, "noblock_validity"}, | |
240799cd | 1360 | {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, |
b3881f74 | 1361 | {Opt_journal_ioprio, "journal_ioprio=%u"}, |
afd4672d | 1362 | {Opt_auto_da_alloc, "auto_da_alloc=%u"}, |
06705bff TT |
1363 | {Opt_auto_da_alloc, "auto_da_alloc"}, |
1364 | {Opt_noauto_da_alloc, "noauto_da_alloc"}, | |
744692dc JZ |
1365 | {Opt_dioread_nolock, "dioread_nolock"}, |
1366 | {Opt_dioread_lock, "dioread_lock"}, | |
5328e635 ES |
1367 | {Opt_discard, "discard"}, |
1368 | {Opt_nodiscard, "nodiscard"}, | |
bfff6873 LC |
1369 | {Opt_init_inode_table, "init_itable=%u"}, |
1370 | {Opt_init_inode_table, "init_itable"}, | |
1371 | {Opt_noinit_inode_table, "noinit_itable"}, | |
f3f12faa | 1372 | {Opt_err, NULL}, |
ac27a0ec DK |
1373 | }; |
1374 | ||
617ba13b | 1375 | static ext4_fsblk_t get_sb_block(void **data) |
ac27a0ec | 1376 | { |
617ba13b | 1377 | ext4_fsblk_t sb_block; |
ac27a0ec DK |
1378 | char *options = (char *) *data; |
1379 | ||
1380 | if (!options || strncmp(options, "sb=", 3) != 0) | |
1381 | return 1; /* Default location */ | |
0b8e58a1 | 1382 | |
ac27a0ec | 1383 | options += 3; |
0b8e58a1 | 1384 | /* TODO: use simple_strtoll with >32bit ext4 */ |
ac27a0ec DK |
1385 | sb_block = simple_strtoul(options, &options, 0); |
1386 | if (*options && *options != ',') { | |
4776004f | 1387 | printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", |
ac27a0ec DK |
1388 | (char *) *data); |
1389 | return 1; | |
1390 | } | |
1391 | if (*options == ',') | |
1392 | options++; | |
1393 | *data = (void *) options; | |
0b8e58a1 | 1394 | |
ac27a0ec DK |
1395 | return sb_block; |
1396 | } | |
1397 | ||
b3881f74 | 1398 | #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) |
437ca0fd DM |
1399 | static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n" |
1400 | "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; | |
b3881f74 | 1401 | |
56c50f11 DM |
1402 | #ifdef CONFIG_QUOTA |
1403 | static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) | |
1404 | { | |
1405 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1406 | char *qname; | |
1407 | ||
1408 | if (sb_any_quota_loaded(sb) && | |
1409 | !sbi->s_qf_names[qtype]) { | |
1410 | ext4_msg(sb, KERN_ERR, | |
1411 | "Cannot change journaled " | |
1412 | "quota options when quota turned on"); | |
1413 | return 0; | |
1414 | } | |
1415 | qname = match_strdup(args); | |
1416 | if (!qname) { | |
1417 | ext4_msg(sb, KERN_ERR, | |
1418 | "Not enough memory for storing quotafile name"); | |
1419 | return 0; | |
1420 | } | |
1421 | if (sbi->s_qf_names[qtype] && | |
1422 | strcmp(sbi->s_qf_names[qtype], qname)) { | |
1423 | ext4_msg(sb, KERN_ERR, | |
1424 | "%s quota file already specified", QTYPE2NAME(qtype)); | |
1425 | kfree(qname); | |
1426 | return 0; | |
1427 | } | |
1428 | sbi->s_qf_names[qtype] = qname; | |
1429 | if (strchr(sbi->s_qf_names[qtype], '/')) { | |
1430 | ext4_msg(sb, KERN_ERR, | |
1431 | "quotafile must be on filesystem root"); | |
1432 | kfree(sbi->s_qf_names[qtype]); | |
1433 | sbi->s_qf_names[qtype] = NULL; | |
1434 | return 0; | |
1435 | } | |
fd8c37ec | 1436 | set_opt(sb, QUOTA); |
56c50f11 DM |
1437 | return 1; |
1438 | } | |
1439 | ||
1440 | static int clear_qf_name(struct super_block *sb, int qtype) | |
1441 | { | |
1442 | ||
1443 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1444 | ||
1445 | if (sb_any_quota_loaded(sb) && | |
1446 | sbi->s_qf_names[qtype]) { | |
1447 | ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" | |
1448 | " when quota turned on"); | |
1449 | return 0; | |
1450 | } | |
1451 | /* | |
1452 | * The space will be released later when all options are confirmed | |
1453 | * to be correct | |
1454 | */ | |
1455 | sbi->s_qf_names[qtype] = NULL; | |
1456 | return 1; | |
1457 | } | |
1458 | #endif | |
1459 | ||
2b2d6d01 | 1460 | static int parse_options(char *options, struct super_block *sb, |
c3191067 | 1461 | unsigned long *journal_devnum, |
b3881f74 | 1462 | unsigned int *journal_ioprio, |
2b2d6d01 | 1463 | ext4_fsblk_t *n_blocks_count, int is_remount) |
ac27a0ec | 1464 | { |
617ba13b | 1465 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
2b2d6d01 | 1466 | char *p; |
ac27a0ec DK |
1467 | substring_t args[MAX_OPT_ARGS]; |
1468 | int data_opt = 0; | |
1469 | int option; | |
1470 | #ifdef CONFIG_QUOTA | |
56c50f11 | 1471 | int qfmt; |
ac27a0ec DK |
1472 | #endif |
1473 | ||
1474 | if (!options) | |
1475 | return 1; | |
1476 | ||
2b2d6d01 | 1477 | while ((p = strsep(&options, ",")) != NULL) { |
ac27a0ec DK |
1478 | int token; |
1479 | if (!*p) | |
1480 | continue; | |
1481 | ||
15121c18 ES |
1482 | /* |
1483 | * Initialize args struct so we know whether arg was | |
1484 | * found; some options take optional arguments. | |
1485 | */ | |
7dc57615 | 1486 | args[0].to = args[0].from = NULL; |
ac27a0ec DK |
1487 | token = match_token(p, tokens, args); |
1488 | switch (token) { | |
1489 | case Opt_bsd_df: | |
437ca0fd | 1490 | ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); |
fd8c37ec | 1491 | clear_opt(sb, MINIX_DF); |
ac27a0ec DK |
1492 | break; |
1493 | case Opt_minix_df: | |
437ca0fd | 1494 | ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); |
fd8c37ec | 1495 | set_opt(sb, MINIX_DF); |
437ca0fd | 1496 | |
ac27a0ec DK |
1497 | break; |
1498 | case Opt_grpid: | |
437ca0fd | 1499 | ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); |
fd8c37ec | 1500 | set_opt(sb, GRPID); |
437ca0fd | 1501 | |
ac27a0ec DK |
1502 | break; |
1503 | case Opt_nogrpid: | |
437ca0fd | 1504 | ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); |
fd8c37ec | 1505 | clear_opt(sb, GRPID); |
437ca0fd | 1506 | |
ac27a0ec DK |
1507 | break; |
1508 | case Opt_resuid: | |
1509 | if (match_int(&args[0], &option)) | |
1510 | return 0; | |
1511 | sbi->s_resuid = option; | |
1512 | break; | |
1513 | case Opt_resgid: | |
1514 | if (match_int(&args[0], &option)) | |
1515 | return 0; | |
1516 | sbi->s_resgid = option; | |
1517 | break; | |
1518 | case Opt_sb: | |
1519 | /* handled by get_sb_block() instead of here */ | |
1520 | /* *sb_block = match_int(&args[0]); */ | |
1521 | break; | |
1522 | case Opt_err_panic: | |
fd8c37ec TT |
1523 | clear_opt(sb, ERRORS_CONT); |
1524 | clear_opt(sb, ERRORS_RO); | |
1525 | set_opt(sb, ERRORS_PANIC); | |
ac27a0ec DK |
1526 | break; |
1527 | case Opt_err_ro: | |
fd8c37ec TT |
1528 | clear_opt(sb, ERRORS_CONT); |
1529 | clear_opt(sb, ERRORS_PANIC); | |
1530 | set_opt(sb, ERRORS_RO); | |
ac27a0ec DK |
1531 | break; |
1532 | case Opt_err_cont: | |
fd8c37ec TT |
1533 | clear_opt(sb, ERRORS_RO); |
1534 | clear_opt(sb, ERRORS_PANIC); | |
1535 | set_opt(sb, ERRORS_CONT); | |
ac27a0ec DK |
1536 | break; |
1537 | case Opt_nouid32: | |
fd8c37ec | 1538 | set_opt(sb, NO_UID32); |
ac27a0ec | 1539 | break; |
ac27a0ec | 1540 | case Opt_debug: |
fd8c37ec | 1541 | set_opt(sb, DEBUG); |
ac27a0ec DK |
1542 | break; |
1543 | case Opt_oldalloc: | |
fd8c37ec | 1544 | set_opt(sb, OLDALLOC); |
ac27a0ec DK |
1545 | break; |
1546 | case Opt_orlov: | |
fd8c37ec | 1547 | clear_opt(sb, OLDALLOC); |
ac27a0ec | 1548 | break; |
03010a33 | 1549 | #ifdef CONFIG_EXT4_FS_XATTR |
ac27a0ec | 1550 | case Opt_user_xattr: |
fd8c37ec | 1551 | set_opt(sb, XATTR_USER); |
ac27a0ec DK |
1552 | break; |
1553 | case Opt_nouser_xattr: | |
fd8c37ec | 1554 | clear_opt(sb, XATTR_USER); |
ac27a0ec DK |
1555 | break; |
1556 | #else | |
1557 | case Opt_user_xattr: | |
1558 | case Opt_nouser_xattr: | |
b31e1552 | 1559 | ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported"); |
ac27a0ec DK |
1560 | break; |
1561 | #endif | |
03010a33 | 1562 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
ac27a0ec | 1563 | case Opt_acl: |
fd8c37ec | 1564 | set_opt(sb, POSIX_ACL); |
ac27a0ec DK |
1565 | break; |
1566 | case Opt_noacl: | |
fd8c37ec | 1567 | clear_opt(sb, POSIX_ACL); |
ac27a0ec DK |
1568 | break; |
1569 | #else | |
1570 | case Opt_acl: | |
1571 | case Opt_noacl: | |
b31e1552 | 1572 | ext4_msg(sb, KERN_ERR, "(no)acl options not supported"); |
ac27a0ec DK |
1573 | break; |
1574 | #endif | |
ac27a0ec DK |
1575 | case Opt_journal_update: |
1576 | /* @@@ FIXME */ | |
1577 | /* Eventually we will want to be able to create | |
1578 | a journal file here. For now, only allow the | |
1579 | user to specify an existing inode to be the | |
1580 | journal file. */ | |
1581 | if (is_remount) { | |
b31e1552 ES |
1582 | ext4_msg(sb, KERN_ERR, |
1583 | "Cannot specify journal on remount"); | |
ac27a0ec DK |
1584 | return 0; |
1585 | } | |
fd8c37ec | 1586 | set_opt(sb, UPDATE_JOURNAL); |
ac27a0ec | 1587 | break; |
ac27a0ec DK |
1588 | case Opt_journal_dev: |
1589 | if (is_remount) { | |
b31e1552 ES |
1590 | ext4_msg(sb, KERN_ERR, |
1591 | "Cannot specify journal on remount"); | |
ac27a0ec DK |
1592 | return 0; |
1593 | } | |
1594 | if (match_int(&args[0], &option)) | |
1595 | return 0; | |
1596 | *journal_devnum = option; | |
1597 | break; | |
818d276c | 1598 | case Opt_journal_checksum: |
fd8c37ec | 1599 | set_opt(sb, JOURNAL_CHECKSUM); |
d4da6c9c | 1600 | break; |
818d276c | 1601 | case Opt_journal_async_commit: |
fd8c37ec TT |
1602 | set_opt(sb, JOURNAL_ASYNC_COMMIT); |
1603 | set_opt(sb, JOURNAL_CHECKSUM); | |
818d276c | 1604 | break; |
ac27a0ec | 1605 | case Opt_noload: |
fd8c37ec | 1606 | set_opt(sb, NOLOAD); |
ac27a0ec DK |
1607 | break; |
1608 | case Opt_commit: | |
1609 | if (match_int(&args[0], &option)) | |
1610 | return 0; | |
1611 | if (option < 0) | |
1612 | return 0; | |
1613 | if (option == 0) | |
cd02ff0b | 1614 | option = JBD2_DEFAULT_MAX_COMMIT_AGE; |
ac27a0ec DK |
1615 | sbi->s_commit_interval = HZ * option; |
1616 | break; | |
30773840 TT |
1617 | case Opt_max_batch_time: |
1618 | if (match_int(&args[0], &option)) | |
1619 | return 0; | |
1620 | if (option < 0) | |
1621 | return 0; | |
1622 | if (option == 0) | |
1623 | option = EXT4_DEF_MAX_BATCH_TIME; | |
1624 | sbi->s_max_batch_time = option; | |
1625 | break; | |
1626 | case Opt_min_batch_time: | |
1627 | if (match_int(&args[0], &option)) | |
1628 | return 0; | |
1629 | if (option < 0) | |
1630 | return 0; | |
1631 | sbi->s_min_batch_time = option; | |
1632 | break; | |
ac27a0ec | 1633 | case Opt_data_journal: |
617ba13b | 1634 | data_opt = EXT4_MOUNT_JOURNAL_DATA; |
ac27a0ec DK |
1635 | goto datacheck; |
1636 | case Opt_data_ordered: | |
617ba13b | 1637 | data_opt = EXT4_MOUNT_ORDERED_DATA; |
ac27a0ec DK |
1638 | goto datacheck; |
1639 | case Opt_data_writeback: | |
617ba13b | 1640 | data_opt = EXT4_MOUNT_WRITEBACK_DATA; |
ac27a0ec DK |
1641 | datacheck: |
1642 | if (is_remount) { | |
482a7425 | 1643 | if (test_opt(sb, DATA_FLAGS) != data_opt) { |
b31e1552 ES |
1644 | ext4_msg(sb, KERN_ERR, |
1645 | "Cannot change data mode on remount"); | |
ac27a0ec DK |
1646 | return 0; |
1647 | } | |
1648 | } else { | |
fd8c37ec | 1649 | clear_opt(sb, DATA_FLAGS); |
ac27a0ec DK |
1650 | sbi->s_mount_opt |= data_opt; |
1651 | } | |
1652 | break; | |
5bf5683a | 1653 | case Opt_data_err_abort: |
fd8c37ec | 1654 | set_opt(sb, DATA_ERR_ABORT); |
5bf5683a HK |
1655 | break; |
1656 | case Opt_data_err_ignore: | |
fd8c37ec | 1657 | clear_opt(sb, DATA_ERR_ABORT); |
5bf5683a | 1658 | break; |
ac27a0ec DK |
1659 | #ifdef CONFIG_QUOTA |
1660 | case Opt_usrjquota: | |
56c50f11 | 1661 | if (!set_qf_name(sb, USRQUOTA, &args[0])) |
ac27a0ec | 1662 | return 0; |
56c50f11 DM |
1663 | break; |
1664 | case Opt_grpjquota: | |
1665 | if (!set_qf_name(sb, GRPQUOTA, &args[0])) | |
ac27a0ec | 1666 | return 0; |
ac27a0ec DK |
1667 | break; |
1668 | case Opt_offusrjquota: | |
56c50f11 DM |
1669 | if (!clear_qf_name(sb, USRQUOTA)) |
1670 | return 0; | |
1671 | break; | |
ac27a0ec | 1672 | case Opt_offgrpjquota: |
56c50f11 | 1673 | if (!clear_qf_name(sb, GRPQUOTA)) |
ac27a0ec | 1674 | return 0; |
ac27a0ec | 1675 | break; |
56c50f11 | 1676 | |
ac27a0ec | 1677 | case Opt_jqfmt_vfsold: |
dfc5d03f JK |
1678 | qfmt = QFMT_VFS_OLD; |
1679 | goto set_qf_format; | |
ac27a0ec | 1680 | case Opt_jqfmt_vfsv0: |
dfc5d03f | 1681 | qfmt = QFMT_VFS_V0; |
5a20bdfc JK |
1682 | goto set_qf_format; |
1683 | case Opt_jqfmt_vfsv1: | |
1684 | qfmt = QFMT_VFS_V1; | |
dfc5d03f | 1685 | set_qf_format: |
17bd13b3 | 1686 | if (sb_any_quota_loaded(sb) && |
dfc5d03f | 1687 | sbi->s_jquota_fmt != qfmt) { |
b31e1552 | 1688 | ext4_msg(sb, KERN_ERR, "Cannot change " |
dfc5d03f | 1689 | "journaled quota options when " |
b31e1552 | 1690 | "quota turned on"); |
dfc5d03f JK |
1691 | return 0; |
1692 | } | |
1693 | sbi->s_jquota_fmt = qfmt; | |
ac27a0ec DK |
1694 | break; |
1695 | case Opt_quota: | |
1696 | case Opt_usrquota: | |
fd8c37ec TT |
1697 | set_opt(sb, QUOTA); |
1698 | set_opt(sb, USRQUOTA); | |
ac27a0ec DK |
1699 | break; |
1700 | case Opt_grpquota: | |
fd8c37ec TT |
1701 | set_opt(sb, QUOTA); |
1702 | set_opt(sb, GRPQUOTA); | |
ac27a0ec DK |
1703 | break; |
1704 | case Opt_noquota: | |
17bd13b3 | 1705 | if (sb_any_quota_loaded(sb)) { |
b31e1552 ES |
1706 | ext4_msg(sb, KERN_ERR, "Cannot change quota " |
1707 | "options when quota turned on"); | |
ac27a0ec DK |
1708 | return 0; |
1709 | } | |
fd8c37ec TT |
1710 | clear_opt(sb, QUOTA); |
1711 | clear_opt(sb, USRQUOTA); | |
1712 | clear_opt(sb, GRPQUOTA); | |
ac27a0ec DK |
1713 | break; |
1714 | #else | |
1715 | case Opt_quota: | |
1716 | case Opt_usrquota: | |
1717 | case Opt_grpquota: | |
b31e1552 ES |
1718 | ext4_msg(sb, KERN_ERR, |
1719 | "quota options not supported"); | |
cd59e7b9 | 1720 | break; |
ac27a0ec DK |
1721 | case Opt_usrjquota: |
1722 | case Opt_grpjquota: | |
1723 | case Opt_offusrjquota: | |
1724 | case Opt_offgrpjquota: | |
1725 | case Opt_jqfmt_vfsold: | |
1726 | case Opt_jqfmt_vfsv0: | |
5a20bdfc | 1727 | case Opt_jqfmt_vfsv1: |
b31e1552 ES |
1728 | ext4_msg(sb, KERN_ERR, |
1729 | "journaled quota options not supported"); | |
ac27a0ec DK |
1730 | break; |
1731 | case Opt_noquota: | |
1732 | break; | |
1733 | #endif | |
1734 | case Opt_abort: | |
4ab2f15b | 1735 | sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; |
ac27a0ec | 1736 | break; |
06705bff | 1737 | case Opt_nobarrier: |
fd8c37ec | 1738 | clear_opt(sb, BARRIER); |
06705bff | 1739 | break; |
ac27a0ec | 1740 | case Opt_barrier: |
15121c18 ES |
1741 | if (args[0].from) { |
1742 | if (match_int(&args[0], &option)) | |
1743 | return 0; | |
1744 | } else | |
1745 | option = 1; /* No argument, default to 1 */ | |
ac27a0ec | 1746 | if (option) |
fd8c37ec | 1747 | set_opt(sb, BARRIER); |
ac27a0ec | 1748 | else |
fd8c37ec | 1749 | clear_opt(sb, BARRIER); |
ac27a0ec DK |
1750 | break; |
1751 | case Opt_ignore: | |
1752 | break; | |
1753 | case Opt_resize: | |
1754 | if (!is_remount) { | |
b31e1552 ES |
1755 | ext4_msg(sb, KERN_ERR, |
1756 | "resize option only available " | |
1757 | "for remount"); | |
ac27a0ec DK |
1758 | return 0; |
1759 | } | |
1760 | if (match_int(&args[0], &option) != 0) | |
1761 | return 0; | |
1762 | *n_blocks_count = option; | |
1763 | break; | |
1764 | case Opt_nobh: | |
206f7ab4 CH |
1765 | ext4_msg(sb, KERN_WARNING, |
1766 | "Ignoring deprecated nobh option"); | |
ac27a0ec DK |
1767 | break; |
1768 | case Opt_bh: | |
206f7ab4 CH |
1769 | ext4_msg(sb, KERN_WARNING, |
1770 | "Ignoring deprecated bh option"); | |
ac27a0ec | 1771 | break; |
25ec56b5 | 1772 | case Opt_i_version: |
fd8c37ec | 1773 | set_opt(sb, I_VERSION); |
25ec56b5 JNC |
1774 | sb->s_flags |= MS_I_VERSION; |
1775 | break; | |
dd919b98 | 1776 | case Opt_nodelalloc: |
fd8c37ec | 1777 | clear_opt(sb, DELALLOC); |
dd919b98 | 1778 | break; |
1449032b | 1779 | case Opt_mblk_io_submit: |
fd8c37ec | 1780 | set_opt(sb, MBLK_IO_SUBMIT); |
1449032b TT |
1781 | break; |
1782 | case Opt_nomblk_io_submit: | |
fd8c37ec | 1783 | clear_opt(sb, MBLK_IO_SUBMIT); |
1449032b | 1784 | break; |
c9de560d AT |
1785 | case Opt_stripe: |
1786 | if (match_int(&args[0], &option)) | |
1787 | return 0; | |
1788 | if (option < 0) | |
1789 | return 0; | |
1790 | sbi->s_stripe = option; | |
1791 | break; | |
64769240 | 1792 | case Opt_delalloc: |
fd8c37ec | 1793 | set_opt(sb, DELALLOC); |
64769240 | 1794 | break; |
6fd058f7 | 1795 | case Opt_block_validity: |
fd8c37ec | 1796 | set_opt(sb, BLOCK_VALIDITY); |
6fd058f7 TT |
1797 | break; |
1798 | case Opt_noblock_validity: | |
fd8c37ec | 1799 | clear_opt(sb, BLOCK_VALIDITY); |
6fd058f7 | 1800 | break; |
240799cd TT |
1801 | case Opt_inode_readahead_blks: |
1802 | if (match_int(&args[0], &option)) | |
1803 | return 0; | |
1804 | if (option < 0 || option > (1 << 30)) | |
1805 | return 0; | |
5dbd571d | 1806 | if (option && !is_power_of_2(option)) { |
b31e1552 ES |
1807 | ext4_msg(sb, KERN_ERR, |
1808 | "EXT4-fs: inode_readahead_blks" | |
1809 | " must be a power of 2"); | |
3197ebdb TT |
1810 | return 0; |
1811 | } | |
240799cd TT |
1812 | sbi->s_inode_readahead_blks = option; |
1813 | break; | |
b3881f74 TT |
1814 | case Opt_journal_ioprio: |
1815 | if (match_int(&args[0], &option)) | |
1816 | return 0; | |
1817 | if (option < 0 || option > 7) | |
1818 | break; | |
1819 | *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, | |
1820 | option); | |
1821 | break; | |
06705bff | 1822 | case Opt_noauto_da_alloc: |
fd8c37ec | 1823 | set_opt(sb, NO_AUTO_DA_ALLOC); |
06705bff | 1824 | break; |
afd4672d | 1825 | case Opt_auto_da_alloc: |
15121c18 ES |
1826 | if (args[0].from) { |
1827 | if (match_int(&args[0], &option)) | |
1828 | return 0; | |
1829 | } else | |
1830 | option = 1; /* No argument, default to 1 */ | |
afd4672d | 1831 | if (option) |
fd8c37ec | 1832 | clear_opt(sb, NO_AUTO_DA_ALLOC); |
afd4672d | 1833 | else |
fd8c37ec | 1834 | set_opt(sb,NO_AUTO_DA_ALLOC); |
afd4672d | 1835 | break; |
5328e635 | 1836 | case Opt_discard: |
fd8c37ec | 1837 | set_opt(sb, DISCARD); |
5328e635 ES |
1838 | break; |
1839 | case Opt_nodiscard: | |
fd8c37ec | 1840 | clear_opt(sb, DISCARD); |
5328e635 | 1841 | break; |
744692dc | 1842 | case Opt_dioread_nolock: |
fd8c37ec | 1843 | set_opt(sb, DIOREAD_NOLOCK); |
744692dc JZ |
1844 | break; |
1845 | case Opt_dioread_lock: | |
fd8c37ec | 1846 | clear_opt(sb, DIOREAD_NOLOCK); |
744692dc | 1847 | break; |
bfff6873 | 1848 | case Opt_init_inode_table: |
fd8c37ec | 1849 | set_opt(sb, INIT_INODE_TABLE); |
bfff6873 LC |
1850 | if (args[0].from) { |
1851 | if (match_int(&args[0], &option)) | |
1852 | return 0; | |
1853 | } else | |
1854 | option = EXT4_DEF_LI_WAIT_MULT; | |
1855 | if (option < 0) | |
1856 | return 0; | |
1857 | sbi->s_li_wait_mult = option; | |
1858 | break; | |
1859 | case Opt_noinit_inode_table: | |
fd8c37ec | 1860 | clear_opt(sb, INIT_INODE_TABLE); |
bfff6873 | 1861 | break; |
ac27a0ec | 1862 | default: |
b31e1552 ES |
1863 | ext4_msg(sb, KERN_ERR, |
1864 | "Unrecognized mount option \"%s\" " | |
1865 | "or missing value", p); | |
ac27a0ec DK |
1866 | return 0; |
1867 | } | |
1868 | } | |
1869 | #ifdef CONFIG_QUOTA | |
1870 | if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { | |
482a7425 | 1871 | if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA]) |
fd8c37ec | 1872 | clear_opt(sb, USRQUOTA); |
ac27a0ec | 1873 | |
482a7425 | 1874 | if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA]) |
fd8c37ec | 1875 | clear_opt(sb, GRPQUOTA); |
ac27a0ec | 1876 | |
56c50f11 | 1877 | if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { |
b31e1552 ES |
1878 | ext4_msg(sb, KERN_ERR, "old and new quota " |
1879 | "format mixing"); | |
ac27a0ec DK |
1880 | return 0; |
1881 | } | |
1882 | ||
1883 | if (!sbi->s_jquota_fmt) { | |
b31e1552 ES |
1884 | ext4_msg(sb, KERN_ERR, "journaled quota format " |
1885 | "not specified"); | |
ac27a0ec DK |
1886 | return 0; |
1887 | } | |
1888 | } else { | |
1889 | if (sbi->s_jquota_fmt) { | |
b31e1552 | 1890 | ext4_msg(sb, KERN_ERR, "journaled quota format " |
2c8be6b2 | 1891 | "specified with no journaling " |
b31e1552 | 1892 | "enabled"); |
ac27a0ec DK |
1893 | return 0; |
1894 | } | |
1895 | } | |
1896 | #endif | |
1897 | return 1; | |
1898 | } | |
1899 | ||
617ba13b | 1900 | static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, |
ac27a0ec DK |
1901 | int read_only) |
1902 | { | |
617ba13b | 1903 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec DK |
1904 | int res = 0; |
1905 | ||
617ba13b | 1906 | if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { |
b31e1552 ES |
1907 | ext4_msg(sb, KERN_ERR, "revision level too high, " |
1908 | "forcing read-only mode"); | |
ac27a0ec DK |
1909 | res = MS_RDONLY; |
1910 | } | |
1911 | if (read_only) | |
1912 | return res; | |
617ba13b | 1913 | if (!(sbi->s_mount_state & EXT4_VALID_FS)) |
b31e1552 ES |
1914 | ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " |
1915 | "running e2fsck is recommended"); | |
617ba13b | 1916 | else if ((sbi->s_mount_state & EXT4_ERROR_FS)) |
b31e1552 ES |
1917 | ext4_msg(sb, KERN_WARNING, |
1918 | "warning: mounting fs with errors, " | |
1919 | "running e2fsck is recommended"); | |
ed3ce80a | 1920 | else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && |
ac27a0ec DK |
1921 | le16_to_cpu(es->s_mnt_count) >= |
1922 | (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) | |
b31e1552 ES |
1923 | ext4_msg(sb, KERN_WARNING, |
1924 | "warning: maximal mount count reached, " | |
1925 | "running e2fsck is recommended"); | |
ac27a0ec DK |
1926 | else if (le32_to_cpu(es->s_checkinterval) && |
1927 | (le32_to_cpu(es->s_lastcheck) + | |
1928 | le32_to_cpu(es->s_checkinterval) <= get_seconds())) | |
b31e1552 ES |
1929 | ext4_msg(sb, KERN_WARNING, |
1930 | "warning: checktime reached, " | |
1931 | "running e2fsck is recommended"); | |
0b8e58a1 | 1932 | if (!sbi->s_journal) |
0390131b | 1933 | es->s_state &= cpu_to_le16(~EXT4_VALID_FS); |
ac27a0ec | 1934 | if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) |
617ba13b | 1935 | es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); |
e8546d06 | 1936 | le16_add_cpu(&es->s_mnt_count, 1); |
ac27a0ec | 1937 | es->s_mtime = cpu_to_le32(get_seconds()); |
617ba13b | 1938 | ext4_update_dynamic_rev(sb); |
0390131b FM |
1939 | if (sbi->s_journal) |
1940 | EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
ac27a0ec | 1941 | |
e2d67052 | 1942 | ext4_commit_super(sb, 1); |
ac27a0ec | 1943 | if (test_opt(sb, DEBUG)) |
a9df9a49 | 1944 | printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " |
a2595b8a | 1945 | "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", |
ac27a0ec DK |
1946 | sb->s_blocksize, |
1947 | sbi->s_groups_count, | |
617ba13b MC |
1948 | EXT4_BLOCKS_PER_GROUP(sb), |
1949 | EXT4_INODES_PER_GROUP(sb), | |
a2595b8a | 1950 | sbi->s_mount_opt, sbi->s_mount_opt2); |
ac27a0ec | 1951 | |
7abc52c2 | 1952 | cleancache_init_fs(sb); |
ac27a0ec DK |
1953 | return res; |
1954 | } | |
1955 | ||
772cb7c8 JS |
1956 | static int ext4_fill_flex_info(struct super_block *sb) |
1957 | { | |
1958 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1959 | struct ext4_group_desc *gdp = NULL; | |
772cb7c8 JS |
1960 | ext4_group_t flex_group_count; |
1961 | ext4_group_t flex_group; | |
1962 | int groups_per_flex = 0; | |
c5ca7c76 | 1963 | size_t size; |
772cb7c8 JS |
1964 | int i; |
1965 | ||
503358ae TT |
1966 | sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; |
1967 | groups_per_flex = 1 << sbi->s_log_groups_per_flex; | |
1968 | ||
1969 | if (groups_per_flex < 2) { | |
772cb7c8 JS |
1970 | sbi->s_log_groups_per_flex = 0; |
1971 | return 1; | |
1972 | } | |
1973 | ||
c62a11fd FB |
1974 | /* We allocate both existing and potentially added groups */ |
1975 | flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) + | |
d94e99a6 AK |
1976 | ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) << |
1977 | EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex; | |
c5ca7c76 TT |
1978 | size = flex_group_count * sizeof(struct flex_groups); |
1979 | sbi->s_flex_groups = kzalloc(size, GFP_KERNEL); | |
1980 | if (sbi->s_flex_groups == NULL) { | |
94de56ab JP |
1981 | sbi->s_flex_groups = vzalloc(size); |
1982 | if (sbi->s_flex_groups == NULL) { | |
1983 | ext4_msg(sb, KERN_ERR, | |
1984 | "not enough memory for %u flex groups", | |
1985 | flex_group_count); | |
1986 | goto failed; | |
1987 | } | |
772cb7c8 | 1988 | } |
772cb7c8 | 1989 | |
772cb7c8 | 1990 | for (i = 0; i < sbi->s_groups_count; i++) { |
88b6edd1 | 1991 | gdp = ext4_get_group_desc(sb, i, NULL); |
772cb7c8 JS |
1992 | |
1993 | flex_group = ext4_flex_group(sbi, i); | |
7ad9bb65 TT |
1994 | atomic_add(ext4_free_inodes_count(sb, gdp), |
1995 | &sbi->s_flex_groups[flex_group].free_inodes); | |
1996 | atomic_add(ext4_free_blks_count(sb, gdp), | |
1997 | &sbi->s_flex_groups[flex_group].free_blocks); | |
1998 | atomic_add(ext4_used_dirs_count(sb, gdp), | |
1999 | &sbi->s_flex_groups[flex_group].used_dirs); | |
772cb7c8 JS |
2000 | } |
2001 | ||
2002 | return 1; | |
2003 | failed: | |
2004 | return 0; | |
2005 | } | |
2006 | ||
717d50e4 AD |
2007 | __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group, |
2008 | struct ext4_group_desc *gdp) | |
2009 | { | |
2010 | __u16 crc = 0; | |
2011 | ||
2012 | if (sbi->s_es->s_feature_ro_compat & | |
2013 | cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) { | |
2014 | int offset = offsetof(struct ext4_group_desc, bg_checksum); | |
2015 | __le32 le_group = cpu_to_le32(block_group); | |
2016 | ||
2017 | crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); | |
2018 | crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); | |
2019 | crc = crc16(crc, (__u8 *)gdp, offset); | |
2020 | offset += sizeof(gdp->bg_checksum); /* skip checksum */ | |
2021 | /* for checksum of struct ext4_group_desc do the rest...*/ | |
2022 | if ((sbi->s_es->s_feature_incompat & | |
2023 | cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) && | |
2024 | offset < le16_to_cpu(sbi->s_es->s_desc_size)) | |
2025 | crc = crc16(crc, (__u8 *)gdp + offset, | |
2026 | le16_to_cpu(sbi->s_es->s_desc_size) - | |
2027 | offset); | |
2028 | } | |
2029 | ||
2030 | return cpu_to_le16(crc); | |
2031 | } | |
2032 | ||
2033 | int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group, | |
2034 | struct ext4_group_desc *gdp) | |
2035 | { | |
2036 | if ((sbi->s_es->s_feature_ro_compat & | |
2037 | cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) && | |
2038 | (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp))) | |
2039 | return 0; | |
2040 | ||
2041 | return 1; | |
2042 | } | |
2043 | ||
ac27a0ec | 2044 | /* Called at mount-time, super-block is locked */ |
bfff6873 LC |
2045 | static int ext4_check_descriptors(struct super_block *sb, |
2046 | ext4_group_t *first_not_zeroed) | |
ac27a0ec | 2047 | { |
617ba13b MC |
2048 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
2049 | ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); | |
2050 | ext4_fsblk_t last_block; | |
bd81d8ee LV |
2051 | ext4_fsblk_t block_bitmap; |
2052 | ext4_fsblk_t inode_bitmap; | |
2053 | ext4_fsblk_t inode_table; | |
ce421581 | 2054 | int flexbg_flag = 0; |
bfff6873 | 2055 | ext4_group_t i, grp = sbi->s_groups_count; |
ac27a0ec | 2056 | |
ce421581 JS |
2057 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) |
2058 | flexbg_flag = 1; | |
2059 | ||
af5bc92d | 2060 | ext4_debug("Checking group descriptors"); |
ac27a0ec | 2061 | |
197cd65a AM |
2062 | for (i = 0; i < sbi->s_groups_count; i++) { |
2063 | struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); | |
2064 | ||
ce421581 | 2065 | if (i == sbi->s_groups_count - 1 || flexbg_flag) |
bd81d8ee | 2066 | last_block = ext4_blocks_count(sbi->s_es) - 1; |
ac27a0ec DK |
2067 | else |
2068 | last_block = first_block + | |
617ba13b | 2069 | (EXT4_BLOCKS_PER_GROUP(sb) - 1); |
ac27a0ec | 2070 | |
bfff6873 LC |
2071 | if ((grp == sbi->s_groups_count) && |
2072 | !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) | |
2073 | grp = i; | |
2074 | ||
8fadc143 | 2075 | block_bitmap = ext4_block_bitmap(sb, gdp); |
2b2d6d01 | 2076 | if (block_bitmap < first_block || block_bitmap > last_block) { |
b31e1552 | 2077 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 2078 | "Block bitmap for group %u not in group " |
b31e1552 | 2079 | "(block %llu)!", i, block_bitmap); |
ac27a0ec DK |
2080 | return 0; |
2081 | } | |
8fadc143 | 2082 | inode_bitmap = ext4_inode_bitmap(sb, gdp); |
2b2d6d01 | 2083 | if (inode_bitmap < first_block || inode_bitmap > last_block) { |
b31e1552 | 2084 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 2085 | "Inode bitmap for group %u not in group " |
b31e1552 | 2086 | "(block %llu)!", i, inode_bitmap); |
ac27a0ec DK |
2087 | return 0; |
2088 | } | |
8fadc143 | 2089 | inode_table = ext4_inode_table(sb, gdp); |
bd81d8ee | 2090 | if (inode_table < first_block || |
2b2d6d01 | 2091 | inode_table + sbi->s_itb_per_group - 1 > last_block) { |
b31e1552 | 2092 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 2093 | "Inode table for group %u not in group " |
b31e1552 | 2094 | "(block %llu)!", i, inode_table); |
ac27a0ec DK |
2095 | return 0; |
2096 | } | |
955ce5f5 | 2097 | ext4_lock_group(sb, i); |
717d50e4 | 2098 | if (!ext4_group_desc_csum_verify(sbi, i, gdp)) { |
b31e1552 ES |
2099 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
2100 | "Checksum for group %u failed (%u!=%u)", | |
2101 | i, le16_to_cpu(ext4_group_desc_csum(sbi, i, | |
2102 | gdp)), le16_to_cpu(gdp->bg_checksum)); | |
7ee1ec4c | 2103 | if (!(sb->s_flags & MS_RDONLY)) { |
955ce5f5 | 2104 | ext4_unlock_group(sb, i); |
8a266467 | 2105 | return 0; |
7ee1ec4c | 2106 | } |
717d50e4 | 2107 | } |
955ce5f5 | 2108 | ext4_unlock_group(sb, i); |
ce421581 JS |
2109 | if (!flexbg_flag) |
2110 | first_block += EXT4_BLOCKS_PER_GROUP(sb); | |
ac27a0ec | 2111 | } |
bfff6873 LC |
2112 | if (NULL != first_not_zeroed) |
2113 | *first_not_zeroed = grp; | |
ac27a0ec | 2114 | |
bd81d8ee | 2115 | ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb)); |
0b8e58a1 | 2116 | sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb)); |
ac27a0ec DK |
2117 | return 1; |
2118 | } | |
2119 | ||
617ba13b | 2120 | /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at |
ac27a0ec DK |
2121 | * the superblock) which were deleted from all directories, but held open by |
2122 | * a process at the time of a crash. We walk the list and try to delete these | |
2123 | * inodes at recovery time (only with a read-write filesystem). | |
2124 | * | |
2125 | * In order to keep the orphan inode chain consistent during traversal (in | |
2126 | * case of crash during recovery), we link each inode into the superblock | |
2127 | * orphan list_head and handle it the same way as an inode deletion during | |
2128 | * normal operation (which journals the operations for us). | |
2129 | * | |
2130 | * We only do an iget() and an iput() on each inode, which is very safe if we | |
2131 | * accidentally point at an in-use or already deleted inode. The worst that | |
2132 | * can happen in this case is that we get a "bit already cleared" message from | |
617ba13b | 2133 | * ext4_free_inode(). The only reason we would point at a wrong inode is if |
ac27a0ec DK |
2134 | * e2fsck was run on this filesystem, and it must have already done the orphan |
2135 | * inode cleanup for us, so we can safely abort without any further action. | |
2136 | */ | |
2b2d6d01 TT |
2137 | static void ext4_orphan_cleanup(struct super_block *sb, |
2138 | struct ext4_super_block *es) | |
ac27a0ec DK |
2139 | { |
2140 | unsigned int s_flags = sb->s_flags; | |
2141 | int nr_orphans = 0, nr_truncates = 0; | |
2142 | #ifdef CONFIG_QUOTA | |
2143 | int i; | |
2144 | #endif | |
2145 | if (!es->s_last_orphan) { | |
2146 | jbd_debug(4, "no orphan inodes to clean up\n"); | |
2147 | return; | |
2148 | } | |
2149 | ||
a8f48a95 | 2150 | if (bdev_read_only(sb->s_bdev)) { |
b31e1552 ES |
2151 | ext4_msg(sb, KERN_ERR, "write access " |
2152 | "unavailable, skipping orphan cleanup"); | |
a8f48a95 ES |
2153 | return; |
2154 | } | |
2155 | ||
d39195c3 AG |
2156 | /* Check if feature set would not allow a r/w mount */ |
2157 | if (!ext4_feature_set_ok(sb, 0)) { | |
2158 | ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " | |
2159 | "unknown ROCOMPAT features"); | |
2160 | return; | |
2161 | } | |
2162 | ||
617ba13b | 2163 | if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { |
ac27a0ec DK |
2164 | if (es->s_last_orphan) |
2165 | jbd_debug(1, "Errors on filesystem, " | |
2166 | "clearing orphan list.\n"); | |
2167 | es->s_last_orphan = 0; | |
2168 | jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); | |
2169 | return; | |
2170 | } | |
2171 | ||
2172 | if (s_flags & MS_RDONLY) { | |
b31e1552 | 2173 | ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); |
ac27a0ec DK |
2174 | sb->s_flags &= ~MS_RDONLY; |
2175 | } | |
2176 | #ifdef CONFIG_QUOTA | |
2177 | /* Needed for iput() to work correctly and not trash data */ | |
2178 | sb->s_flags |= MS_ACTIVE; | |
2179 | /* Turn on quotas so that they are updated correctly */ | |
2180 | for (i = 0; i < MAXQUOTAS; i++) { | |
617ba13b MC |
2181 | if (EXT4_SB(sb)->s_qf_names[i]) { |
2182 | int ret = ext4_quota_on_mount(sb, i); | |
ac27a0ec | 2183 | if (ret < 0) |
b31e1552 ES |
2184 | ext4_msg(sb, KERN_ERR, |
2185 | "Cannot turn on journaled " | |
2186 | "quota: error %d", ret); | |
ac27a0ec DK |
2187 | } |
2188 | } | |
2189 | #endif | |
2190 | ||
2191 | while (es->s_last_orphan) { | |
2192 | struct inode *inode; | |
2193 | ||
97bd42b9 JB |
2194 | inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); |
2195 | if (IS_ERR(inode)) { | |
ac27a0ec DK |
2196 | es->s_last_orphan = 0; |
2197 | break; | |
2198 | } | |
2199 | ||
617ba13b | 2200 | list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); |
871a2931 | 2201 | dquot_initialize(inode); |
ac27a0ec | 2202 | if (inode->i_nlink) { |
b31e1552 ES |
2203 | ext4_msg(sb, KERN_DEBUG, |
2204 | "%s: truncating inode %lu to %lld bytes", | |
46e665e9 | 2205 | __func__, inode->i_ino, inode->i_size); |
e5f8eab8 | 2206 | jbd_debug(2, "truncating inode %lu to %lld bytes\n", |
ac27a0ec | 2207 | inode->i_ino, inode->i_size); |
617ba13b | 2208 | ext4_truncate(inode); |
ac27a0ec DK |
2209 | nr_truncates++; |
2210 | } else { | |
b31e1552 ES |
2211 | ext4_msg(sb, KERN_DEBUG, |
2212 | "%s: deleting unreferenced inode %lu", | |
46e665e9 | 2213 | __func__, inode->i_ino); |
ac27a0ec DK |
2214 | jbd_debug(2, "deleting unreferenced inode %lu\n", |
2215 | inode->i_ino); | |
2216 | nr_orphans++; | |
2217 | } | |
2218 | iput(inode); /* The delete magic happens here! */ | |
2219 | } | |
2220 | ||
2b2d6d01 | 2221 | #define PLURAL(x) (x), ((x) == 1) ? "" : "s" |
ac27a0ec DK |
2222 | |
2223 | if (nr_orphans) | |
b31e1552 ES |
2224 | ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", |
2225 | PLURAL(nr_orphans)); | |
ac27a0ec | 2226 | if (nr_truncates) |
b31e1552 ES |
2227 | ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", |
2228 | PLURAL(nr_truncates)); | |
ac27a0ec DK |
2229 | #ifdef CONFIG_QUOTA |
2230 | /* Turn quotas off */ | |
2231 | for (i = 0; i < MAXQUOTAS; i++) { | |
2232 | if (sb_dqopt(sb)->files[i]) | |
287a8095 | 2233 | dquot_quota_off(sb, i); |
ac27a0ec DK |
2234 | } |
2235 | #endif | |
2236 | sb->s_flags = s_flags; /* Restore MS_RDONLY status */ | |
2237 | } | |
0b8e58a1 | 2238 | |
cd2291a4 ES |
2239 | /* |
2240 | * Maximal extent format file size. | |
2241 | * Resulting logical blkno at s_maxbytes must fit in our on-disk | |
2242 | * extent format containers, within a sector_t, and within i_blocks | |
2243 | * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, | |
2244 | * so that won't be a limiting factor. | |
2245 | * | |
f17722f9 LC |
2246 | * However there is other limiting factor. We do store extents in the form |
2247 | * of starting block and length, hence the resulting length of the extent | |
2248 | * covering maximum file size must fit into on-disk format containers as | |
2249 | * well. Given that length is always by 1 unit bigger than max unit (because | |
2250 | * we count 0 as well) we have to lower the s_maxbytes by one fs block. | |
2251 | * | |
cd2291a4 ES |
2252 | * Note, this does *not* consider any metadata overhead for vfs i_blocks. |
2253 | */ | |
f287a1a5 | 2254 | static loff_t ext4_max_size(int blkbits, int has_huge_files) |
cd2291a4 ES |
2255 | { |
2256 | loff_t res; | |
2257 | loff_t upper_limit = MAX_LFS_FILESIZE; | |
2258 | ||
2259 | /* small i_blocks in vfs inode? */ | |
f287a1a5 | 2260 | if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { |
cd2291a4 | 2261 | /* |
90c699a9 | 2262 | * CONFIG_LBDAF is not enabled implies the inode |
cd2291a4 ES |
2263 | * i_block represent total blocks in 512 bytes |
2264 | * 32 == size of vfs inode i_blocks * 8 | |
2265 | */ | |
2266 | upper_limit = (1LL << 32) - 1; | |
2267 | ||
2268 | /* total blocks in file system block size */ | |
2269 | upper_limit >>= (blkbits - 9); | |
2270 | upper_limit <<= blkbits; | |
2271 | } | |
2272 | ||
f17722f9 LC |
2273 | /* |
2274 | * 32-bit extent-start container, ee_block. We lower the maxbytes | |
2275 | * by one fs block, so ee_len can cover the extent of maximum file | |
2276 | * size | |
2277 | */ | |
2278 | res = (1LL << 32) - 1; | |
cd2291a4 | 2279 | res <<= blkbits; |
cd2291a4 ES |
2280 | |
2281 | /* Sanity check against vm- & vfs- imposed limits */ | |
2282 | if (res > upper_limit) | |
2283 | res = upper_limit; | |
2284 | ||
2285 | return res; | |
2286 | } | |
ac27a0ec | 2287 | |
ac27a0ec | 2288 | /* |
cd2291a4 | 2289 | * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect |
0fc1b451 AK |
2290 | * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. |
2291 | * We need to be 1 filesystem block less than the 2^48 sector limit. | |
ac27a0ec | 2292 | */ |
f287a1a5 | 2293 | static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) |
ac27a0ec | 2294 | { |
617ba13b | 2295 | loff_t res = EXT4_NDIR_BLOCKS; |
0fc1b451 AK |
2296 | int meta_blocks; |
2297 | loff_t upper_limit; | |
0b8e58a1 AD |
2298 | /* This is calculated to be the largest file size for a dense, block |
2299 | * mapped file such that the file's total number of 512-byte sectors, | |
2300 | * including data and all indirect blocks, does not exceed (2^48 - 1). | |
2301 | * | |
2302 | * __u32 i_blocks_lo and _u16 i_blocks_high represent the total | |
2303 | * number of 512-byte sectors of the file. | |
0fc1b451 AK |
2304 | */ |
2305 | ||
f287a1a5 | 2306 | if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { |
0fc1b451 | 2307 | /* |
90c699a9 | 2308 | * !has_huge_files or CONFIG_LBDAF not enabled implies that |
0b8e58a1 AD |
2309 | * the inode i_block field represents total file blocks in |
2310 | * 2^32 512-byte sectors == size of vfs inode i_blocks * 8 | |
0fc1b451 AK |
2311 | */ |
2312 | upper_limit = (1LL << 32) - 1; | |
2313 | ||
2314 | /* total blocks in file system block size */ | |
2315 | upper_limit >>= (bits - 9); | |
2316 | ||
2317 | } else { | |
8180a562 AK |
2318 | /* |
2319 | * We use 48 bit ext4_inode i_blocks | |
2320 | * With EXT4_HUGE_FILE_FL set the i_blocks | |
2321 | * represent total number of blocks in | |
2322 | * file system block size | |
2323 | */ | |
0fc1b451 AK |
2324 | upper_limit = (1LL << 48) - 1; |
2325 | ||
0fc1b451 AK |
2326 | } |
2327 | ||
2328 | /* indirect blocks */ | |
2329 | meta_blocks = 1; | |
2330 | /* double indirect blocks */ | |
2331 | meta_blocks += 1 + (1LL << (bits-2)); | |
2332 | /* tripple indirect blocks */ | |
2333 | meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); | |
2334 | ||
2335 | upper_limit -= meta_blocks; | |
2336 | upper_limit <<= bits; | |
ac27a0ec DK |
2337 | |
2338 | res += 1LL << (bits-2); | |
2339 | res += 1LL << (2*(bits-2)); | |
2340 | res += 1LL << (3*(bits-2)); | |
2341 | res <<= bits; | |
2342 | if (res > upper_limit) | |
2343 | res = upper_limit; | |
0fc1b451 AK |
2344 | |
2345 | if (res > MAX_LFS_FILESIZE) | |
2346 | res = MAX_LFS_FILESIZE; | |
2347 | ||
ac27a0ec DK |
2348 | return res; |
2349 | } | |
2350 | ||
617ba13b | 2351 | static ext4_fsblk_t descriptor_loc(struct super_block *sb, |
0b8e58a1 | 2352 | ext4_fsblk_t logical_sb_block, int nr) |
ac27a0ec | 2353 | { |
617ba13b | 2354 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
fd2d4291 | 2355 | ext4_group_t bg, first_meta_bg; |
ac27a0ec DK |
2356 | int has_super = 0; |
2357 | ||
2358 | first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); | |
2359 | ||
617ba13b | 2360 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) || |
ac27a0ec | 2361 | nr < first_meta_bg) |
70bbb3e0 | 2362 | return logical_sb_block + nr + 1; |
ac27a0ec | 2363 | bg = sbi->s_desc_per_block * nr; |
617ba13b | 2364 | if (ext4_bg_has_super(sb, bg)) |
ac27a0ec | 2365 | has_super = 1; |
0b8e58a1 | 2366 | |
617ba13b | 2367 | return (has_super + ext4_group_first_block_no(sb, bg)); |
ac27a0ec DK |
2368 | } |
2369 | ||
c9de560d AT |
2370 | /** |
2371 | * ext4_get_stripe_size: Get the stripe size. | |
2372 | * @sbi: In memory super block info | |
2373 | * | |
2374 | * If we have specified it via mount option, then | |
2375 | * use the mount option value. If the value specified at mount time is | |
2376 | * greater than the blocks per group use the super block value. | |
2377 | * If the super block value is greater than blocks per group return 0. | |
2378 | * Allocator needs it be less than blocks per group. | |
2379 | * | |
2380 | */ | |
2381 | static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) | |
2382 | { | |
2383 | unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); | |
2384 | unsigned long stripe_width = | |
2385 | le32_to_cpu(sbi->s_es->s_raid_stripe_width); | |
2386 | ||
2387 | if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) | |
2388 | return sbi->s_stripe; | |
2389 | ||
2390 | if (stripe_width <= sbi->s_blocks_per_group) | |
2391 | return stripe_width; | |
2392 | ||
2393 | if (stride <= sbi->s_blocks_per_group) | |
2394 | return stride; | |
2395 | ||
2396 | return 0; | |
2397 | } | |
ac27a0ec | 2398 | |
3197ebdb TT |
2399 | /* sysfs supprt */ |
2400 | ||
2401 | struct ext4_attr { | |
2402 | struct attribute attr; | |
2403 | ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *); | |
60e6679e | 2404 | ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *, |
3197ebdb TT |
2405 | const char *, size_t); |
2406 | int offset; | |
2407 | }; | |
2408 | ||
2409 | static int parse_strtoul(const char *buf, | |
2410 | unsigned long max, unsigned long *value) | |
2411 | { | |
2412 | char *endp; | |
2413 | ||
e7d2860b AGR |
2414 | *value = simple_strtoul(skip_spaces(buf), &endp, 0); |
2415 | endp = skip_spaces(endp); | |
3197ebdb TT |
2416 | if (*endp || *value > max) |
2417 | return -EINVAL; | |
2418 | ||
2419 | return 0; | |
2420 | } | |
2421 | ||
2422 | static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a, | |
2423 | struct ext4_sb_info *sbi, | |
2424 | char *buf) | |
2425 | { | |
2426 | return snprintf(buf, PAGE_SIZE, "%llu\n", | |
2427 | (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter)); | |
2428 | } | |
2429 | ||
2430 | static ssize_t session_write_kbytes_show(struct ext4_attr *a, | |
2431 | struct ext4_sb_info *sbi, char *buf) | |
2432 | { | |
2433 | struct super_block *sb = sbi->s_buddy_cache->i_sb; | |
2434 | ||
f613dfcb TT |
2435 | if (!sb->s_bdev->bd_part) |
2436 | return snprintf(buf, PAGE_SIZE, "0\n"); | |
3197ebdb TT |
2437 | return snprintf(buf, PAGE_SIZE, "%lu\n", |
2438 | (part_stat_read(sb->s_bdev->bd_part, sectors[1]) - | |
2439 | sbi->s_sectors_written_start) >> 1); | |
2440 | } | |
2441 | ||
2442 | static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a, | |
2443 | struct ext4_sb_info *sbi, char *buf) | |
2444 | { | |
2445 | struct super_block *sb = sbi->s_buddy_cache->i_sb; | |
2446 | ||
f613dfcb TT |
2447 | if (!sb->s_bdev->bd_part) |
2448 | return snprintf(buf, PAGE_SIZE, "0\n"); | |
3197ebdb | 2449 | return snprintf(buf, PAGE_SIZE, "%llu\n", |
a6b43e38 | 2450 | (unsigned long long)(sbi->s_kbytes_written + |
3197ebdb | 2451 | ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - |
a6b43e38 | 2452 | EXT4_SB(sb)->s_sectors_written_start) >> 1))); |
3197ebdb TT |
2453 | } |
2454 | ||
77f4135f VH |
2455 | static ssize_t extent_cache_hits_show(struct ext4_attr *a, |
2456 | struct ext4_sb_info *sbi, char *buf) | |
2457 | { | |
2458 | return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits); | |
2459 | } | |
2460 | ||
2461 | static ssize_t extent_cache_misses_show(struct ext4_attr *a, | |
2462 | struct ext4_sb_info *sbi, char *buf) | |
2463 | { | |
2464 | return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses); | |
2465 | } | |
2466 | ||
3197ebdb TT |
2467 | static ssize_t inode_readahead_blks_store(struct ext4_attr *a, |
2468 | struct ext4_sb_info *sbi, | |
2469 | const char *buf, size_t count) | |
2470 | { | |
2471 | unsigned long t; | |
2472 | ||
2473 | if (parse_strtoul(buf, 0x40000000, &t)) | |
2474 | return -EINVAL; | |
2475 | ||
5dbd571d | 2476 | if (t && !is_power_of_2(t)) |
3197ebdb TT |
2477 | return -EINVAL; |
2478 | ||
2479 | sbi->s_inode_readahead_blks = t; | |
2480 | return count; | |
2481 | } | |
2482 | ||
2483 | static ssize_t sbi_ui_show(struct ext4_attr *a, | |
0b8e58a1 | 2484 | struct ext4_sb_info *sbi, char *buf) |
3197ebdb TT |
2485 | { |
2486 | unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset); | |
2487 | ||
2488 | return snprintf(buf, PAGE_SIZE, "%u\n", *ui); | |
2489 | } | |
2490 | ||
2491 | static ssize_t sbi_ui_store(struct ext4_attr *a, | |
2492 | struct ext4_sb_info *sbi, | |
2493 | const char *buf, size_t count) | |
2494 | { | |
2495 | unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset); | |
2496 | unsigned long t; | |
2497 | ||
2498 | if (parse_strtoul(buf, 0xffffffff, &t)) | |
2499 | return -EINVAL; | |
2500 | *ui = t; | |
2501 | return count; | |
2502 | } | |
2503 | ||
2504 | #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \ | |
2505 | static struct ext4_attr ext4_attr_##_name = { \ | |
2506 | .attr = {.name = __stringify(_name), .mode = _mode }, \ | |
2507 | .show = _show, \ | |
2508 | .store = _store, \ | |
2509 | .offset = offsetof(struct ext4_sb_info, _elname), \ | |
2510 | } | |
2511 | #define EXT4_ATTR(name, mode, show, store) \ | |
2512 | static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store) | |
2513 | ||
857ac889 | 2514 | #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL) |
3197ebdb TT |
2515 | #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL) |
2516 | #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store) | |
2517 | #define EXT4_RW_ATTR_SBI_UI(name, elname) \ | |
2518 | EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname) | |
2519 | #define ATTR_LIST(name) &ext4_attr_##name.attr | |
2520 | ||
2521 | EXT4_RO_ATTR(delayed_allocation_blocks); | |
2522 | EXT4_RO_ATTR(session_write_kbytes); | |
2523 | EXT4_RO_ATTR(lifetime_write_kbytes); | |
77f4135f VH |
2524 | EXT4_RO_ATTR(extent_cache_hits); |
2525 | EXT4_RO_ATTR(extent_cache_misses); | |
3197ebdb TT |
2526 | EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show, |
2527 | inode_readahead_blks_store, s_inode_readahead_blks); | |
11013911 | 2528 | EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal); |
3197ebdb TT |
2529 | EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats); |
2530 | EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan); | |
2531 | EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan); | |
2532 | EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs); | |
2533 | EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request); | |
2534 | EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc); | |
55138e0b | 2535 | EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump); |
3197ebdb TT |
2536 | |
2537 | static struct attribute *ext4_attrs[] = { | |
2538 | ATTR_LIST(delayed_allocation_blocks), | |
2539 | ATTR_LIST(session_write_kbytes), | |
2540 | ATTR_LIST(lifetime_write_kbytes), | |
77f4135f VH |
2541 | ATTR_LIST(extent_cache_hits), |
2542 | ATTR_LIST(extent_cache_misses), | |
3197ebdb | 2543 | ATTR_LIST(inode_readahead_blks), |
11013911 | 2544 | ATTR_LIST(inode_goal), |
3197ebdb TT |
2545 | ATTR_LIST(mb_stats), |
2546 | ATTR_LIST(mb_max_to_scan), | |
2547 | ATTR_LIST(mb_min_to_scan), | |
2548 | ATTR_LIST(mb_order2_req), | |
2549 | ATTR_LIST(mb_stream_req), | |
2550 | ATTR_LIST(mb_group_prealloc), | |
55138e0b | 2551 | ATTR_LIST(max_writeback_mb_bump), |
3197ebdb TT |
2552 | NULL, |
2553 | }; | |
2554 | ||
857ac889 LC |
2555 | /* Features this copy of ext4 supports */ |
2556 | EXT4_INFO_ATTR(lazy_itable_init); | |
27ee40df | 2557 | EXT4_INFO_ATTR(batched_discard); |
857ac889 LC |
2558 | |
2559 | static struct attribute *ext4_feat_attrs[] = { | |
2560 | ATTR_LIST(lazy_itable_init), | |
27ee40df | 2561 | ATTR_LIST(batched_discard), |
857ac889 LC |
2562 | NULL, |
2563 | }; | |
2564 | ||
3197ebdb TT |
2565 | static ssize_t ext4_attr_show(struct kobject *kobj, |
2566 | struct attribute *attr, char *buf) | |
2567 | { | |
2568 | struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, | |
2569 | s_kobj); | |
2570 | struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); | |
2571 | ||
2572 | return a->show ? a->show(a, sbi, buf) : 0; | |
2573 | } | |
2574 | ||
2575 | static ssize_t ext4_attr_store(struct kobject *kobj, | |
2576 | struct attribute *attr, | |
2577 | const char *buf, size_t len) | |
2578 | { | |
2579 | struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, | |
2580 | s_kobj); | |
2581 | struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); | |
2582 | ||
2583 | return a->store ? a->store(a, sbi, buf, len) : 0; | |
2584 | } | |
2585 | ||
2586 | static void ext4_sb_release(struct kobject *kobj) | |
2587 | { | |
2588 | struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, | |
2589 | s_kobj); | |
2590 | complete(&sbi->s_kobj_unregister); | |
2591 | } | |
2592 | ||
52cf25d0 | 2593 | static const struct sysfs_ops ext4_attr_ops = { |
3197ebdb TT |
2594 | .show = ext4_attr_show, |
2595 | .store = ext4_attr_store, | |
2596 | }; | |
2597 | ||
2598 | static struct kobj_type ext4_ktype = { | |
2599 | .default_attrs = ext4_attrs, | |
2600 | .sysfs_ops = &ext4_attr_ops, | |
2601 | .release = ext4_sb_release, | |
2602 | }; | |
2603 | ||
857ac889 LC |
2604 | static void ext4_feat_release(struct kobject *kobj) |
2605 | { | |
2606 | complete(&ext4_feat->f_kobj_unregister); | |
2607 | } | |
2608 | ||
2609 | static struct kobj_type ext4_feat_ktype = { | |
2610 | .default_attrs = ext4_feat_attrs, | |
2611 | .sysfs_ops = &ext4_attr_ops, | |
2612 | .release = ext4_feat_release, | |
2613 | }; | |
2614 | ||
a13fb1a4 ES |
2615 | /* |
2616 | * Check whether this filesystem can be mounted based on | |
2617 | * the features present and the RDONLY/RDWR mount requested. | |
2618 | * Returns 1 if this filesystem can be mounted as requested, | |
2619 | * 0 if it cannot be. | |
2620 | */ | |
2621 | static int ext4_feature_set_ok(struct super_block *sb, int readonly) | |
2622 | { | |
2623 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) { | |
2624 | ext4_msg(sb, KERN_ERR, | |
2625 | "Couldn't mount because of " | |
2626 | "unsupported optional features (%x)", | |
2627 | (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & | |
2628 | ~EXT4_FEATURE_INCOMPAT_SUPP)); | |
2629 | return 0; | |
2630 | } | |
2631 | ||
2632 | if (readonly) | |
2633 | return 1; | |
2634 | ||
2635 | /* Check that feature set is OK for a read-write mount */ | |
2636 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) { | |
2637 | ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " | |
2638 | "unsupported optional features (%x)", | |
2639 | (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & | |
2640 | ~EXT4_FEATURE_RO_COMPAT_SUPP)); | |
2641 | return 0; | |
2642 | } | |
2643 | /* | |
2644 | * Large file size enabled file system can only be mounted | |
2645 | * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF | |
2646 | */ | |
2647 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) { | |
2648 | if (sizeof(blkcnt_t) < sizeof(u64)) { | |
2649 | ext4_msg(sb, KERN_ERR, "Filesystem with huge files " | |
2650 | "cannot be mounted RDWR without " | |
2651 | "CONFIG_LBDAF"); | |
2652 | return 0; | |
2653 | } | |
2654 | } | |
2655 | return 1; | |
2656 | } | |
2657 | ||
66e61a9e TT |
2658 | /* |
2659 | * This function is called once a day if we have errors logged | |
2660 | * on the file system | |
2661 | */ | |
2662 | static void print_daily_error_info(unsigned long arg) | |
2663 | { | |
2664 | struct super_block *sb = (struct super_block *) arg; | |
2665 | struct ext4_sb_info *sbi; | |
2666 | struct ext4_super_block *es; | |
2667 | ||
2668 | sbi = EXT4_SB(sb); | |
2669 | es = sbi->s_es; | |
2670 | ||
2671 | if (es->s_error_count) | |
2672 | ext4_msg(sb, KERN_NOTICE, "error count: %u", | |
2673 | le32_to_cpu(es->s_error_count)); | |
2674 | if (es->s_first_error_time) { | |
2675 | printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d", | |
2676 | sb->s_id, le32_to_cpu(es->s_first_error_time), | |
2677 | (int) sizeof(es->s_first_error_func), | |
2678 | es->s_first_error_func, | |
2679 | le32_to_cpu(es->s_first_error_line)); | |
2680 | if (es->s_first_error_ino) | |
2681 | printk(": inode %u", | |
2682 | le32_to_cpu(es->s_first_error_ino)); | |
2683 | if (es->s_first_error_block) | |
2684 | printk(": block %llu", (unsigned long long) | |
2685 | le64_to_cpu(es->s_first_error_block)); | |
2686 | printk("\n"); | |
2687 | } | |
2688 | if (es->s_last_error_time) { | |
2689 | printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d", | |
2690 | sb->s_id, le32_to_cpu(es->s_last_error_time), | |
2691 | (int) sizeof(es->s_last_error_func), | |
2692 | es->s_last_error_func, | |
2693 | le32_to_cpu(es->s_last_error_line)); | |
2694 | if (es->s_last_error_ino) | |
2695 | printk(": inode %u", | |
2696 | le32_to_cpu(es->s_last_error_ino)); | |
2697 | if (es->s_last_error_block) | |
2698 | printk(": block %llu", (unsigned long long) | |
2699 | le64_to_cpu(es->s_last_error_block)); | |
2700 | printk("\n"); | |
2701 | } | |
2702 | mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ | |
2703 | } | |
2704 | ||
bfff6873 LC |
2705 | /* Find next suitable group and run ext4_init_inode_table */ |
2706 | static int ext4_run_li_request(struct ext4_li_request *elr) | |
2707 | { | |
2708 | struct ext4_group_desc *gdp = NULL; | |
2709 | ext4_group_t group, ngroups; | |
2710 | struct super_block *sb; | |
2711 | unsigned long timeout = 0; | |
2712 | int ret = 0; | |
2713 | ||
2714 | sb = elr->lr_super; | |
2715 | ngroups = EXT4_SB(sb)->s_groups_count; | |
2716 | ||
2717 | for (group = elr->lr_next_group; group < ngroups; group++) { | |
2718 | gdp = ext4_get_group_desc(sb, group, NULL); | |
2719 | if (!gdp) { | |
2720 | ret = 1; | |
2721 | break; | |
2722 | } | |
2723 | ||
2724 | if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) | |
2725 | break; | |
2726 | } | |
2727 | ||
2728 | if (group == ngroups) | |
2729 | ret = 1; | |
2730 | ||
2731 | if (!ret) { | |
2732 | timeout = jiffies; | |
2733 | ret = ext4_init_inode_table(sb, group, | |
2734 | elr->lr_timeout ? 0 : 1); | |
2735 | if (elr->lr_timeout == 0) { | |
51ce6511 LC |
2736 | timeout = (jiffies - timeout) * |
2737 | elr->lr_sbi->s_li_wait_mult; | |
bfff6873 LC |
2738 | elr->lr_timeout = timeout; |
2739 | } | |
2740 | elr->lr_next_sched = jiffies + elr->lr_timeout; | |
2741 | elr->lr_next_group = group + 1; | |
2742 | } | |
2743 | ||
2744 | return ret; | |
2745 | } | |
2746 | ||
2747 | /* | |
2748 | * Remove lr_request from the list_request and free the | |
4ed5c033 | 2749 | * request structure. Should be called with li_list_mtx held |
bfff6873 LC |
2750 | */ |
2751 | static void ext4_remove_li_request(struct ext4_li_request *elr) | |
2752 | { | |
2753 | struct ext4_sb_info *sbi; | |
2754 | ||
2755 | if (!elr) | |
2756 | return; | |
2757 | ||
2758 | sbi = elr->lr_sbi; | |
2759 | ||
2760 | list_del(&elr->lr_request); | |
2761 | sbi->s_li_request = NULL; | |
2762 | kfree(elr); | |
2763 | } | |
2764 | ||
2765 | static void ext4_unregister_li_request(struct super_block *sb) | |
2766 | { | |
1bb933fb LC |
2767 | mutex_lock(&ext4_li_mtx); |
2768 | if (!ext4_li_info) { | |
2769 | mutex_unlock(&ext4_li_mtx); | |
bfff6873 | 2770 | return; |
1bb933fb | 2771 | } |
bfff6873 LC |
2772 | |
2773 | mutex_lock(&ext4_li_info->li_list_mtx); | |
1bb933fb | 2774 | ext4_remove_li_request(EXT4_SB(sb)->s_li_request); |
bfff6873 | 2775 | mutex_unlock(&ext4_li_info->li_list_mtx); |
1bb933fb | 2776 | mutex_unlock(&ext4_li_mtx); |
bfff6873 LC |
2777 | } |
2778 | ||
8f1f7453 ES |
2779 | static struct task_struct *ext4_lazyinit_task; |
2780 | ||
bfff6873 LC |
2781 | /* |
2782 | * This is the function where ext4lazyinit thread lives. It walks | |
2783 | * through the request list searching for next scheduled filesystem. | |
2784 | * When such a fs is found, run the lazy initialization request | |
2785 | * (ext4_rn_li_request) and keep track of the time spend in this | |
2786 | * function. Based on that time we compute next schedule time of | |
2787 | * the request. When walking through the list is complete, compute | |
2788 | * next waking time and put itself into sleep. | |
2789 | */ | |
2790 | static int ext4_lazyinit_thread(void *arg) | |
2791 | { | |
2792 | struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; | |
2793 | struct list_head *pos, *n; | |
2794 | struct ext4_li_request *elr; | |
4ed5c033 | 2795 | unsigned long next_wakeup, cur; |
bfff6873 LC |
2796 | |
2797 | BUG_ON(NULL == eli); | |
2798 | ||
bfff6873 LC |
2799 | cont_thread: |
2800 | while (true) { | |
2801 | next_wakeup = MAX_JIFFY_OFFSET; | |
2802 | ||
2803 | mutex_lock(&eli->li_list_mtx); | |
2804 | if (list_empty(&eli->li_request_list)) { | |
2805 | mutex_unlock(&eli->li_list_mtx); | |
2806 | goto exit_thread; | |
2807 | } | |
2808 | ||
2809 | list_for_each_safe(pos, n, &eli->li_request_list) { | |
2810 | elr = list_entry(pos, struct ext4_li_request, | |
2811 | lr_request); | |
2812 | ||
b2c78cd0 TT |
2813 | if (time_after_eq(jiffies, elr->lr_next_sched)) { |
2814 | if (ext4_run_li_request(elr) != 0) { | |
2815 | /* error, remove the lazy_init job */ | |
2816 | ext4_remove_li_request(elr); | |
2817 | continue; | |
2818 | } | |
bfff6873 LC |
2819 | } |
2820 | ||
2821 | if (time_before(elr->lr_next_sched, next_wakeup)) | |
2822 | next_wakeup = elr->lr_next_sched; | |
2823 | } | |
2824 | mutex_unlock(&eli->li_list_mtx); | |
2825 | ||
2826 | if (freezing(current)) | |
2827 | refrigerator(); | |
2828 | ||
4ed5c033 LC |
2829 | cur = jiffies; |
2830 | if ((time_after_eq(cur, next_wakeup)) || | |
f4245bd4 | 2831 | (MAX_JIFFY_OFFSET == next_wakeup)) { |
bfff6873 LC |
2832 | cond_resched(); |
2833 | continue; | |
2834 | } | |
2835 | ||
4ed5c033 LC |
2836 | schedule_timeout_interruptible(next_wakeup - cur); |
2837 | ||
8f1f7453 ES |
2838 | if (kthread_should_stop()) { |
2839 | ext4_clear_request_list(); | |
2840 | goto exit_thread; | |
2841 | } | |
bfff6873 LC |
2842 | } |
2843 | ||
2844 | exit_thread: | |
2845 | /* | |
2846 | * It looks like the request list is empty, but we need | |
2847 | * to check it under the li_list_mtx lock, to prevent any | |
2848 | * additions into it, and of course we should lock ext4_li_mtx | |
2849 | * to atomically free the list and ext4_li_info, because at | |
2850 | * this point another ext4 filesystem could be registering | |
2851 | * new one. | |
2852 | */ | |
2853 | mutex_lock(&ext4_li_mtx); | |
2854 | mutex_lock(&eli->li_list_mtx); | |
2855 | if (!list_empty(&eli->li_request_list)) { | |
2856 | mutex_unlock(&eli->li_list_mtx); | |
2857 | mutex_unlock(&ext4_li_mtx); | |
2858 | goto cont_thread; | |
2859 | } | |
2860 | mutex_unlock(&eli->li_list_mtx); | |
bfff6873 LC |
2861 | kfree(ext4_li_info); |
2862 | ext4_li_info = NULL; | |
2863 | mutex_unlock(&ext4_li_mtx); | |
2864 | ||
2865 | return 0; | |
2866 | } | |
2867 | ||
2868 | static void ext4_clear_request_list(void) | |
2869 | { | |
2870 | struct list_head *pos, *n; | |
2871 | struct ext4_li_request *elr; | |
2872 | ||
2873 | mutex_lock(&ext4_li_info->li_list_mtx); | |
bfff6873 LC |
2874 | list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { |
2875 | elr = list_entry(pos, struct ext4_li_request, | |
2876 | lr_request); | |
2877 | ext4_remove_li_request(elr); | |
2878 | } | |
2879 | mutex_unlock(&ext4_li_info->li_list_mtx); | |
2880 | } | |
2881 | ||
2882 | static int ext4_run_lazyinit_thread(void) | |
2883 | { | |
8f1f7453 ES |
2884 | ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, |
2885 | ext4_li_info, "ext4lazyinit"); | |
2886 | if (IS_ERR(ext4_lazyinit_task)) { | |
2887 | int err = PTR_ERR(ext4_lazyinit_task); | |
bfff6873 | 2888 | ext4_clear_request_list(); |
bfff6873 LC |
2889 | kfree(ext4_li_info); |
2890 | ext4_li_info = NULL; | |
2891 | printk(KERN_CRIT "EXT4: error %d creating inode table " | |
2892 | "initialization thread\n", | |
2893 | err); | |
2894 | return err; | |
2895 | } | |
2896 | ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; | |
bfff6873 LC |
2897 | return 0; |
2898 | } | |
2899 | ||
2900 | /* | |
2901 | * Check whether it make sense to run itable init. thread or not. | |
2902 | * If there is at least one uninitialized inode table, return | |
2903 | * corresponding group number, else the loop goes through all | |
2904 | * groups and return total number of groups. | |
2905 | */ | |
2906 | static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) | |
2907 | { | |
2908 | ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; | |
2909 | struct ext4_group_desc *gdp = NULL; | |
2910 | ||
2911 | for (group = 0; group < ngroups; group++) { | |
2912 | gdp = ext4_get_group_desc(sb, group, NULL); | |
2913 | if (!gdp) | |
2914 | continue; | |
2915 | ||
2916 | if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) | |
2917 | break; | |
2918 | } | |
2919 | ||
2920 | return group; | |
2921 | } | |
2922 | ||
2923 | static int ext4_li_info_new(void) | |
2924 | { | |
2925 | struct ext4_lazy_init *eli = NULL; | |
2926 | ||
2927 | eli = kzalloc(sizeof(*eli), GFP_KERNEL); | |
2928 | if (!eli) | |
2929 | return -ENOMEM; | |
2930 | ||
bfff6873 LC |
2931 | INIT_LIST_HEAD(&eli->li_request_list); |
2932 | mutex_init(&eli->li_list_mtx); | |
2933 | ||
bfff6873 LC |
2934 | eli->li_state |= EXT4_LAZYINIT_QUIT; |
2935 | ||
2936 | ext4_li_info = eli; | |
2937 | ||
2938 | return 0; | |
2939 | } | |
2940 | ||
2941 | static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, | |
2942 | ext4_group_t start) | |
2943 | { | |
2944 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
2945 | struct ext4_li_request *elr; | |
2946 | unsigned long rnd; | |
2947 | ||
2948 | elr = kzalloc(sizeof(*elr), GFP_KERNEL); | |
2949 | if (!elr) | |
2950 | return NULL; | |
2951 | ||
2952 | elr->lr_super = sb; | |
2953 | elr->lr_sbi = sbi; | |
2954 | elr->lr_next_group = start; | |
2955 | ||
2956 | /* | |
2957 | * Randomize first schedule time of the request to | |
2958 | * spread the inode table initialization requests | |
2959 | * better. | |
2960 | */ | |
2961 | get_random_bytes(&rnd, sizeof(rnd)); | |
2962 | elr->lr_next_sched = jiffies + (unsigned long)rnd % | |
2963 | (EXT4_DEF_LI_MAX_START_DELAY * HZ); | |
2964 | ||
2965 | return elr; | |
2966 | } | |
2967 | ||
2968 | static int ext4_register_li_request(struct super_block *sb, | |
2969 | ext4_group_t first_not_zeroed) | |
2970 | { | |
2971 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
2972 | struct ext4_li_request *elr; | |
2973 | ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; | |
6c5a6cb9 | 2974 | int ret = 0; |
bfff6873 | 2975 | |
51ce6511 LC |
2976 | if (sbi->s_li_request != NULL) { |
2977 | /* | |
2978 | * Reset timeout so it can be computed again, because | |
2979 | * s_li_wait_mult might have changed. | |
2980 | */ | |
2981 | sbi->s_li_request->lr_timeout = 0; | |
beed5ecb | 2982 | return 0; |
51ce6511 | 2983 | } |
bfff6873 LC |
2984 | |
2985 | if (first_not_zeroed == ngroups || | |
2986 | (sb->s_flags & MS_RDONLY) || | |
55ff3840 | 2987 | !test_opt(sb, INIT_INODE_TABLE)) |
beed5ecb | 2988 | return 0; |
bfff6873 LC |
2989 | |
2990 | elr = ext4_li_request_new(sb, first_not_zeroed); | |
beed5ecb NK |
2991 | if (!elr) |
2992 | return -ENOMEM; | |
bfff6873 LC |
2993 | |
2994 | mutex_lock(&ext4_li_mtx); | |
2995 | ||
2996 | if (NULL == ext4_li_info) { | |
2997 | ret = ext4_li_info_new(); | |
2998 | if (ret) | |
2999 | goto out; | |
3000 | } | |
3001 | ||
3002 | mutex_lock(&ext4_li_info->li_list_mtx); | |
3003 | list_add(&elr->lr_request, &ext4_li_info->li_request_list); | |
3004 | mutex_unlock(&ext4_li_info->li_list_mtx); | |
3005 | ||
3006 | sbi->s_li_request = elr; | |
46e4690b TM |
3007 | /* |
3008 | * set elr to NULL here since it has been inserted to | |
3009 | * the request_list and the removal and free of it is | |
3010 | * handled by ext4_clear_request_list from now on. | |
3011 | */ | |
3012 | elr = NULL; | |
bfff6873 LC |
3013 | |
3014 | if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { | |
3015 | ret = ext4_run_lazyinit_thread(); | |
3016 | if (ret) | |
3017 | goto out; | |
3018 | } | |
bfff6873 | 3019 | out: |
beed5ecb NK |
3020 | mutex_unlock(&ext4_li_mtx); |
3021 | if (ret) | |
bfff6873 | 3022 | kfree(elr); |
bfff6873 LC |
3023 | return ret; |
3024 | } | |
3025 | ||
3026 | /* | |
3027 | * We do not need to lock anything since this is called on | |
3028 | * module unload. | |
3029 | */ | |
3030 | static void ext4_destroy_lazyinit_thread(void) | |
3031 | { | |
3032 | /* | |
3033 | * If thread exited earlier | |
3034 | * there's nothing to be done. | |
3035 | */ | |
8f1f7453 | 3036 | if (!ext4_li_info || !ext4_lazyinit_task) |
bfff6873 LC |
3037 | return; |
3038 | ||
8f1f7453 | 3039 | kthread_stop(ext4_lazyinit_task); |
bfff6873 LC |
3040 | } |
3041 | ||
2b2d6d01 | 3042 | static int ext4_fill_super(struct super_block *sb, void *data, int silent) |
7477827f AK |
3043 | __releases(kernel_lock) |
3044 | __acquires(kernel_lock) | |
ac27a0ec | 3045 | { |
d4c402d9 | 3046 | char *orig_data = kstrdup(data, GFP_KERNEL); |
2b2d6d01 | 3047 | struct buffer_head *bh; |
617ba13b MC |
3048 | struct ext4_super_block *es = NULL; |
3049 | struct ext4_sb_info *sbi; | |
3050 | ext4_fsblk_t block; | |
3051 | ext4_fsblk_t sb_block = get_sb_block(&data); | |
70bbb3e0 | 3052 | ext4_fsblk_t logical_sb_block; |
ac27a0ec | 3053 | unsigned long offset = 0; |
ac27a0ec DK |
3054 | unsigned long journal_devnum = 0; |
3055 | unsigned long def_mount_opts; | |
3056 | struct inode *root; | |
9f6200bb | 3057 | char *cp; |
0390131b | 3058 | const char *descr; |
dcc7dae3 | 3059 | int ret = -ENOMEM; |
ac27a0ec | 3060 | int blocksize; |
4ec11028 TT |
3061 | unsigned int db_count; |
3062 | unsigned int i; | |
f287a1a5 | 3063 | int needs_recovery, has_huge_files; |
bd81d8ee | 3064 | __u64 blocks_count; |
833f4077 | 3065 | int err; |
b3881f74 | 3066 | unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; |
bfff6873 | 3067 | ext4_group_t first_not_zeroed; |
ac27a0ec DK |
3068 | |
3069 | sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); | |
3070 | if (!sbi) | |
dcc7dae3 | 3071 | goto out_free_orig; |
705895b6 PE |
3072 | |
3073 | sbi->s_blockgroup_lock = | |
3074 | kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); | |
3075 | if (!sbi->s_blockgroup_lock) { | |
3076 | kfree(sbi); | |
dcc7dae3 | 3077 | goto out_free_orig; |
705895b6 | 3078 | } |
ac27a0ec DK |
3079 | sb->s_fs_info = sbi; |
3080 | sbi->s_mount_opt = 0; | |
617ba13b MC |
3081 | sbi->s_resuid = EXT4_DEF_RESUID; |
3082 | sbi->s_resgid = EXT4_DEF_RESGID; | |
240799cd | 3083 | sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; |
d9c9bef1 | 3084 | sbi->s_sb_block = sb_block; |
f613dfcb TT |
3085 | if (sb->s_bdev->bd_part) |
3086 | sbi->s_sectors_written_start = | |
3087 | part_stat_read(sb->s_bdev->bd_part, sectors[1]); | |
ac27a0ec | 3088 | |
9f6200bb TT |
3089 | /* Cleanup superblock name */ |
3090 | for (cp = sb->s_id; (cp = strchr(cp, '/'));) | |
3091 | *cp = '!'; | |
3092 | ||
dcc7dae3 | 3093 | ret = -EINVAL; |
617ba13b | 3094 | blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); |
ac27a0ec | 3095 | if (!blocksize) { |
b31e1552 | 3096 | ext4_msg(sb, KERN_ERR, "unable to set blocksize"); |
ac27a0ec DK |
3097 | goto out_fail; |
3098 | } | |
3099 | ||
3100 | /* | |
617ba13b | 3101 | * The ext4 superblock will not be buffer aligned for other than 1kB |
ac27a0ec DK |
3102 | * block sizes. We need to calculate the offset from buffer start. |
3103 | */ | |
617ba13b | 3104 | if (blocksize != EXT4_MIN_BLOCK_SIZE) { |
70bbb3e0 AM |
3105 | logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; |
3106 | offset = do_div(logical_sb_block, blocksize); | |
ac27a0ec | 3107 | } else { |
70bbb3e0 | 3108 | logical_sb_block = sb_block; |
ac27a0ec DK |
3109 | } |
3110 | ||
70bbb3e0 | 3111 | if (!(bh = sb_bread(sb, logical_sb_block))) { |
b31e1552 | 3112 | ext4_msg(sb, KERN_ERR, "unable to read superblock"); |
ac27a0ec DK |
3113 | goto out_fail; |
3114 | } | |
3115 | /* | |
3116 | * Note: s_es must be initialized as soon as possible because | |
617ba13b | 3117 | * some ext4 macro-instructions depend on its value |
ac27a0ec | 3118 | */ |
617ba13b | 3119 | es = (struct ext4_super_block *) (((char *)bh->b_data) + offset); |
ac27a0ec DK |
3120 | sbi->s_es = es; |
3121 | sb->s_magic = le16_to_cpu(es->s_magic); | |
617ba13b MC |
3122 | if (sb->s_magic != EXT4_SUPER_MAGIC) |
3123 | goto cantfind_ext4; | |
afc32f7e | 3124 | sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); |
ac27a0ec DK |
3125 | |
3126 | /* Set defaults before we parse the mount options */ | |
3127 | def_mount_opts = le32_to_cpu(es->s_default_mount_opts); | |
fd8c37ec | 3128 | set_opt(sb, INIT_INODE_TABLE); |
617ba13b | 3129 | if (def_mount_opts & EXT4_DEFM_DEBUG) |
fd8c37ec | 3130 | set_opt(sb, DEBUG); |
437ca0fd DM |
3131 | if (def_mount_opts & EXT4_DEFM_BSDGROUPS) { |
3132 | ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups", | |
3133 | "2.6.38"); | |
fd8c37ec | 3134 | set_opt(sb, GRPID); |
437ca0fd | 3135 | } |
617ba13b | 3136 | if (def_mount_opts & EXT4_DEFM_UID16) |
fd8c37ec | 3137 | set_opt(sb, NO_UID32); |
ea663336 | 3138 | /* xattr user namespace & acls are now defaulted on */ |
03010a33 | 3139 | #ifdef CONFIG_EXT4_FS_XATTR |
ea663336 | 3140 | set_opt(sb, XATTR_USER); |
2e7842b8 | 3141 | #endif |
03010a33 | 3142 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
ea663336 | 3143 | set_opt(sb, POSIX_ACL); |
2e7842b8 | 3144 | #endif |
6fd7a467 | 3145 | set_opt(sb, MBLK_IO_SUBMIT); |
617ba13b | 3146 | if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) |
fd8c37ec | 3147 | set_opt(sb, JOURNAL_DATA); |
617ba13b | 3148 | else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) |
fd8c37ec | 3149 | set_opt(sb, ORDERED_DATA); |
617ba13b | 3150 | else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) |
fd8c37ec | 3151 | set_opt(sb, WRITEBACK_DATA); |
617ba13b MC |
3152 | |
3153 | if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) | |
fd8c37ec | 3154 | set_opt(sb, ERRORS_PANIC); |
bb4f397a | 3155 | else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) |
fd8c37ec | 3156 | set_opt(sb, ERRORS_CONT); |
bb4f397a | 3157 | else |
fd8c37ec | 3158 | set_opt(sb, ERRORS_RO); |
8b67f04a | 3159 | if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY) |
fd8c37ec | 3160 | set_opt(sb, BLOCK_VALIDITY); |
8b67f04a | 3161 | if (def_mount_opts & EXT4_DEFM_DISCARD) |
fd8c37ec | 3162 | set_opt(sb, DISCARD); |
ac27a0ec DK |
3163 | |
3164 | sbi->s_resuid = le16_to_cpu(es->s_def_resuid); | |
3165 | sbi->s_resgid = le16_to_cpu(es->s_def_resgid); | |
30773840 TT |
3166 | sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; |
3167 | sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; | |
3168 | sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; | |
ac27a0ec | 3169 | |
8b67f04a | 3170 | if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) |
fd8c37ec | 3171 | set_opt(sb, BARRIER); |
ac27a0ec | 3172 | |
dd919b98 AK |
3173 | /* |
3174 | * enable delayed allocation by default | |
3175 | * Use -o nodelalloc to turn it off | |
3176 | */ | |
8b67f04a TT |
3177 | if (!IS_EXT3_SB(sb) && |
3178 | ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) | |
fd8c37ec | 3179 | set_opt(sb, DELALLOC); |
dd919b98 | 3180 | |
51ce6511 LC |
3181 | /* |
3182 | * set default s_li_wait_mult for lazyinit, for the case there is | |
3183 | * no mount option specified. | |
3184 | */ | |
3185 | sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; | |
3186 | ||
8b67f04a TT |
3187 | if (!parse_options((char *) sbi->s_es->s_mount_opts, sb, |
3188 | &journal_devnum, &journal_ioprio, NULL, 0)) { | |
3189 | ext4_msg(sb, KERN_WARNING, | |
3190 | "failed to parse options in superblock: %s", | |
3191 | sbi->s_es->s_mount_opts); | |
3192 | } | |
b3881f74 TT |
3193 | if (!parse_options((char *) data, sb, &journal_devnum, |
3194 | &journal_ioprio, NULL, 0)) | |
ac27a0ec DK |
3195 | goto failed_mount; |
3196 | ||
3197 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
482a7425 | 3198 | (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); |
ac27a0ec | 3199 | |
617ba13b MC |
3200 | if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && |
3201 | (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) || | |
3202 | EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) || | |
3203 | EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U))) | |
b31e1552 ES |
3204 | ext4_msg(sb, KERN_WARNING, |
3205 | "feature flags set on rev 0 fs, " | |
3206 | "running e2fsck is recommended"); | |
469108ff | 3207 | |
2035e776 TT |
3208 | if (IS_EXT2_SB(sb)) { |
3209 | if (ext2_feature_set_ok(sb)) | |
3210 | ext4_msg(sb, KERN_INFO, "mounting ext2 file system " | |
3211 | "using the ext4 subsystem"); | |
3212 | else { | |
3213 | ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " | |
3214 | "to feature incompatibilities"); | |
3215 | goto failed_mount; | |
3216 | } | |
3217 | } | |
3218 | ||
3219 | if (IS_EXT3_SB(sb)) { | |
3220 | if (ext3_feature_set_ok(sb)) | |
3221 | ext4_msg(sb, KERN_INFO, "mounting ext3 file system " | |
3222 | "using the ext4 subsystem"); | |
3223 | else { | |
3224 | ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " | |
3225 | "to feature incompatibilities"); | |
3226 | goto failed_mount; | |
3227 | } | |
3228 | } | |
3229 | ||
ac27a0ec DK |
3230 | /* |
3231 | * Check feature flags regardless of the revision level, since we | |
3232 | * previously didn't change the revision level when setting the flags, | |
3233 | * so there is a chance incompat flags are set on a rev 0 filesystem. | |
3234 | */ | |
a13fb1a4 | 3235 | if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY))) |
ac27a0ec | 3236 | goto failed_mount; |
a13fb1a4 | 3237 | |
ac27a0ec DK |
3238 | blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); |
3239 | ||
617ba13b MC |
3240 | if (blocksize < EXT4_MIN_BLOCK_SIZE || |
3241 | blocksize > EXT4_MAX_BLOCK_SIZE) { | |
b31e1552 ES |
3242 | ext4_msg(sb, KERN_ERR, |
3243 | "Unsupported filesystem blocksize %d", blocksize); | |
ac27a0ec DK |
3244 | goto failed_mount; |
3245 | } | |
3246 | ||
ac27a0ec | 3247 | if (sb->s_blocksize != blocksize) { |
ce40733c AK |
3248 | /* Validate the filesystem blocksize */ |
3249 | if (!sb_set_blocksize(sb, blocksize)) { | |
b31e1552 | 3250 | ext4_msg(sb, KERN_ERR, "bad block size %d", |
ce40733c | 3251 | blocksize); |
ac27a0ec DK |
3252 | goto failed_mount; |
3253 | } | |
3254 | ||
2b2d6d01 | 3255 | brelse(bh); |
70bbb3e0 AM |
3256 | logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; |
3257 | offset = do_div(logical_sb_block, blocksize); | |
3258 | bh = sb_bread(sb, logical_sb_block); | |
ac27a0ec | 3259 | if (!bh) { |
b31e1552 ES |
3260 | ext4_msg(sb, KERN_ERR, |
3261 | "Can't read superblock on 2nd try"); | |
ac27a0ec DK |
3262 | goto failed_mount; |
3263 | } | |
617ba13b | 3264 | es = (struct ext4_super_block *)(((char *)bh->b_data) + offset); |
ac27a0ec | 3265 | sbi->s_es = es; |
617ba13b | 3266 | if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { |
b31e1552 ES |
3267 | ext4_msg(sb, KERN_ERR, |
3268 | "Magic mismatch, very weird!"); | |
ac27a0ec DK |
3269 | goto failed_mount; |
3270 | } | |
3271 | } | |
3272 | ||
a13fb1a4 ES |
3273 | has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb, |
3274 | EXT4_FEATURE_RO_COMPAT_HUGE_FILE); | |
f287a1a5 TT |
3275 | sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, |
3276 | has_huge_files); | |
3277 | sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); | |
ac27a0ec | 3278 | |
617ba13b MC |
3279 | if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { |
3280 | sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; | |
3281 | sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; | |
ac27a0ec DK |
3282 | } else { |
3283 | sbi->s_inode_size = le16_to_cpu(es->s_inode_size); | |
3284 | sbi->s_first_ino = le32_to_cpu(es->s_first_ino); | |
617ba13b | 3285 | if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || |
1330593e | 3286 | (!is_power_of_2(sbi->s_inode_size)) || |
ac27a0ec | 3287 | (sbi->s_inode_size > blocksize)) { |
b31e1552 ES |
3288 | ext4_msg(sb, KERN_ERR, |
3289 | "unsupported inode size: %d", | |
2b2d6d01 | 3290 | sbi->s_inode_size); |
ac27a0ec DK |
3291 | goto failed_mount; |
3292 | } | |
ef7f3835 KS |
3293 | if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) |
3294 | sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2); | |
ac27a0ec | 3295 | } |
0b8e58a1 | 3296 | |
0d1ee42f AR |
3297 | sbi->s_desc_size = le16_to_cpu(es->s_desc_size); |
3298 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) { | |
8fadc143 | 3299 | if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || |
0d1ee42f | 3300 | sbi->s_desc_size > EXT4_MAX_DESC_SIZE || |
d8ea6cf8 | 3301 | !is_power_of_2(sbi->s_desc_size)) { |
b31e1552 ES |
3302 | ext4_msg(sb, KERN_ERR, |
3303 | "unsupported descriptor size %lu", | |
0d1ee42f AR |
3304 | sbi->s_desc_size); |
3305 | goto failed_mount; | |
3306 | } | |
3307 | } else | |
3308 | sbi->s_desc_size = EXT4_MIN_DESC_SIZE; | |
0b8e58a1 | 3309 | |
ac27a0ec | 3310 | sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); |
ac27a0ec | 3311 | sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); |
b47b6f38 | 3312 | if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0) |
617ba13b | 3313 | goto cantfind_ext4; |
0b8e58a1 | 3314 | |
617ba13b | 3315 | sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); |
ac27a0ec | 3316 | if (sbi->s_inodes_per_block == 0) |
617ba13b | 3317 | goto cantfind_ext4; |
ac27a0ec DK |
3318 | sbi->s_itb_per_group = sbi->s_inodes_per_group / |
3319 | sbi->s_inodes_per_block; | |
0d1ee42f | 3320 | sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); |
ac27a0ec DK |
3321 | sbi->s_sbh = bh; |
3322 | sbi->s_mount_state = le16_to_cpu(es->s_state); | |
e57aa839 FW |
3323 | sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); |
3324 | sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); | |
0b8e58a1 | 3325 | |
2b2d6d01 | 3326 | for (i = 0; i < 4; i++) |
ac27a0ec DK |
3327 | sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); |
3328 | sbi->s_def_hash_version = es->s_def_hash_version; | |
f99b2589 TT |
3329 | i = le32_to_cpu(es->s_flags); |
3330 | if (i & EXT2_FLAGS_UNSIGNED_HASH) | |
3331 | sbi->s_hash_unsigned = 3; | |
3332 | else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { | |
3333 | #ifdef __CHAR_UNSIGNED__ | |
3334 | es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); | |
3335 | sbi->s_hash_unsigned = 3; | |
3336 | #else | |
3337 | es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); | |
3338 | #endif | |
3339 | sb->s_dirt = 1; | |
3340 | } | |
ac27a0ec DK |
3341 | |
3342 | if (sbi->s_blocks_per_group > blocksize * 8) { | |
b31e1552 ES |
3343 | ext4_msg(sb, KERN_ERR, |
3344 | "#blocks per group too big: %lu", | |
2b2d6d01 | 3345 | sbi->s_blocks_per_group); |
ac27a0ec DK |
3346 | goto failed_mount; |
3347 | } | |
ac27a0ec | 3348 | if (sbi->s_inodes_per_group > blocksize * 8) { |
b31e1552 ES |
3349 | ext4_msg(sb, KERN_ERR, |
3350 | "#inodes per group too big: %lu", | |
2b2d6d01 | 3351 | sbi->s_inodes_per_group); |
ac27a0ec DK |
3352 | goto failed_mount; |
3353 | } | |
3354 | ||
bf43d84b ES |
3355 | /* |
3356 | * Test whether we have more sectors than will fit in sector_t, | |
3357 | * and whether the max offset is addressable by the page cache. | |
3358 | */ | |
5a9ae68a | 3359 | err = generic_check_addressable(sb->s_blocksize_bits, |
30ca22c7 | 3360 | ext4_blocks_count(es)); |
5a9ae68a | 3361 | if (err) { |
b31e1552 | 3362 | ext4_msg(sb, KERN_ERR, "filesystem" |
bf43d84b | 3363 | " too large to mount safely on this system"); |
ac27a0ec | 3364 | if (sizeof(sector_t) < 8) |
90c699a9 | 3365 | ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled"); |
5a9ae68a | 3366 | ret = err; |
ac27a0ec DK |
3367 | goto failed_mount; |
3368 | } | |
3369 | ||
617ba13b MC |
3370 | if (EXT4_BLOCKS_PER_GROUP(sb) == 0) |
3371 | goto cantfind_ext4; | |
e7c95593 | 3372 | |
0f2ddca6 FTN |
3373 | /* check blocks count against device size */ |
3374 | blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; | |
3375 | if (blocks_count && ext4_blocks_count(es) > blocks_count) { | |
b31e1552 ES |
3376 | ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " |
3377 | "exceeds size of device (%llu blocks)", | |
0f2ddca6 FTN |
3378 | ext4_blocks_count(es), blocks_count); |
3379 | goto failed_mount; | |
3380 | } | |
3381 | ||
0b8e58a1 AD |
3382 | /* |
3383 | * It makes no sense for the first data block to be beyond the end | |
3384 | * of the filesystem. | |
3385 | */ | |
3386 | if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { | |
b31e1552 ES |
3387 | ext4_msg(sb, KERN_WARNING, "bad geometry: first data" |
3388 | "block %u is beyond end of filesystem (%llu)", | |
3389 | le32_to_cpu(es->s_first_data_block), | |
3390 | ext4_blocks_count(es)); | |
e7c95593 ES |
3391 | goto failed_mount; |
3392 | } | |
bd81d8ee LV |
3393 | blocks_count = (ext4_blocks_count(es) - |
3394 | le32_to_cpu(es->s_first_data_block) + | |
3395 | EXT4_BLOCKS_PER_GROUP(sb) - 1); | |
3396 | do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); | |
4ec11028 | 3397 | if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { |
b31e1552 | 3398 | ext4_msg(sb, KERN_WARNING, "groups count too large: %u " |
4ec11028 | 3399 | "(block count %llu, first data block %u, " |
b31e1552 | 3400 | "blocks per group %lu)", sbi->s_groups_count, |
4ec11028 TT |
3401 | ext4_blocks_count(es), |
3402 | le32_to_cpu(es->s_first_data_block), | |
3403 | EXT4_BLOCKS_PER_GROUP(sb)); | |
3404 | goto failed_mount; | |
3405 | } | |
bd81d8ee | 3406 | sbi->s_groups_count = blocks_count; |
fb0a387d ES |
3407 | sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, |
3408 | (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); | |
617ba13b MC |
3409 | db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / |
3410 | EXT4_DESC_PER_BLOCK(sb); | |
2b2d6d01 | 3411 | sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *), |
ac27a0ec DK |
3412 | GFP_KERNEL); |
3413 | if (sbi->s_group_desc == NULL) { | |
b31e1552 | 3414 | ext4_msg(sb, KERN_ERR, "not enough memory"); |
ac27a0ec DK |
3415 | goto failed_mount; |
3416 | } | |
3417 | ||
3244fcb1 | 3418 | #ifdef CONFIG_PROC_FS |
9f6200bb TT |
3419 | if (ext4_proc_root) |
3420 | sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root); | |
3244fcb1 | 3421 | #endif |
240799cd | 3422 | |
705895b6 | 3423 | bgl_lock_init(sbi->s_blockgroup_lock); |
ac27a0ec DK |
3424 | |
3425 | for (i = 0; i < db_count; i++) { | |
70bbb3e0 | 3426 | block = descriptor_loc(sb, logical_sb_block, i); |
ac27a0ec DK |
3427 | sbi->s_group_desc[i] = sb_bread(sb, block); |
3428 | if (!sbi->s_group_desc[i]) { | |
b31e1552 ES |
3429 | ext4_msg(sb, KERN_ERR, |
3430 | "can't read group descriptor %d", i); | |
ac27a0ec DK |
3431 | db_count = i; |
3432 | goto failed_mount2; | |
3433 | } | |
3434 | } | |
bfff6873 | 3435 | if (!ext4_check_descriptors(sb, &first_not_zeroed)) { |
b31e1552 | 3436 | ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); |
ac27a0ec DK |
3437 | goto failed_mount2; |
3438 | } | |
772cb7c8 JS |
3439 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) |
3440 | if (!ext4_fill_flex_info(sb)) { | |
b31e1552 ES |
3441 | ext4_msg(sb, KERN_ERR, |
3442 | "unable to initialize " | |
3443 | "flex_bg meta info!"); | |
772cb7c8 JS |
3444 | goto failed_mount2; |
3445 | } | |
3446 | ||
ac27a0ec DK |
3447 | sbi->s_gdb_count = db_count; |
3448 | get_random_bytes(&sbi->s_next_generation, sizeof(u32)); | |
3449 | spin_lock_init(&sbi->s_next_gen_lock); | |
3450 | ||
04496411 TM |
3451 | init_timer(&sbi->s_err_report); |
3452 | sbi->s_err_report.function = print_daily_error_info; | |
3453 | sbi->s_err_report.data = (unsigned long) sb; | |
3454 | ||
ce7e010a TT |
3455 | err = percpu_counter_init(&sbi->s_freeblocks_counter, |
3456 | ext4_count_free_blocks(sb)); | |
3457 | if (!err) { | |
3458 | err = percpu_counter_init(&sbi->s_freeinodes_counter, | |
3459 | ext4_count_free_inodes(sb)); | |
3460 | } | |
3461 | if (!err) { | |
3462 | err = percpu_counter_init(&sbi->s_dirs_counter, | |
3463 | ext4_count_dirs(sb)); | |
3464 | } | |
3465 | if (!err) { | |
3466 | err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0); | |
3467 | } | |
3468 | if (err) { | |
3469 | ext4_msg(sb, KERN_ERR, "insufficient memory"); | |
3470 | goto failed_mount3; | |
3471 | } | |
3472 | ||
c9de560d | 3473 | sbi->s_stripe = ext4_get_stripe_size(sbi); |
55138e0b | 3474 | sbi->s_max_writeback_mb_bump = 128; |
c9de560d | 3475 | |
ac27a0ec DK |
3476 | /* |
3477 | * set up enough so that it can read an inode | |
3478 | */ | |
9ca92389 TT |
3479 | if (!test_opt(sb, NOLOAD) && |
3480 | EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) | |
3481 | sb->s_op = &ext4_sops; | |
3482 | else | |
3483 | sb->s_op = &ext4_nojournal_sops; | |
617ba13b MC |
3484 | sb->s_export_op = &ext4_export_ops; |
3485 | sb->s_xattr = ext4_xattr_handlers; | |
ac27a0ec | 3486 | #ifdef CONFIG_QUOTA |
617ba13b MC |
3487 | sb->s_qcop = &ext4_qctl_operations; |
3488 | sb->dq_op = &ext4_quota_operations; | |
ac27a0ec | 3489 | #endif |
f2fa2ffc AK |
3490 | memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); |
3491 | ||
ac27a0ec | 3492 | INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ |
3b9d4ed2 | 3493 | mutex_init(&sbi->s_orphan_lock); |
32ed5058 | 3494 | mutex_init(&sbi->s_resize_lock); |
ac27a0ec DK |
3495 | |
3496 | sb->s_root = NULL; | |
3497 | ||
3498 | needs_recovery = (es->s_last_orphan != 0 || | |
617ba13b MC |
3499 | EXT4_HAS_INCOMPAT_FEATURE(sb, |
3500 | EXT4_FEATURE_INCOMPAT_RECOVER)); | |
ac27a0ec | 3501 | |
c5e06d10 JL |
3502 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) && |
3503 | !(sb->s_flags & MS_RDONLY)) | |
3504 | if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) | |
3505 | goto failed_mount3; | |
3506 | ||
ac27a0ec DK |
3507 | /* |
3508 | * The first inode we look at is the journal inode. Don't try | |
3509 | * root first: it may be modified in the journal! | |
3510 | */ | |
3511 | if (!test_opt(sb, NOLOAD) && | |
617ba13b MC |
3512 | EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { |
3513 | if (ext4_load_journal(sb, es, journal_devnum)) | |
ac27a0ec | 3514 | goto failed_mount3; |
0390131b FM |
3515 | } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) && |
3516 | EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { | |
b31e1552 ES |
3517 | ext4_msg(sb, KERN_ERR, "required journal recovery " |
3518 | "suppressed and not mounted read-only"); | |
744692dc | 3519 | goto failed_mount_wq; |
ac27a0ec | 3520 | } else { |
fd8c37ec | 3521 | clear_opt(sb, DATA_FLAGS); |
0390131b FM |
3522 | sbi->s_journal = NULL; |
3523 | needs_recovery = 0; | |
3524 | goto no_journal; | |
ac27a0ec DK |
3525 | } |
3526 | ||
eb40a09c JS |
3527 | if (ext4_blocks_count(es) > 0xffffffffULL && |
3528 | !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, | |
3529 | JBD2_FEATURE_INCOMPAT_64BIT)) { | |
b31e1552 | 3530 | ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); |
744692dc | 3531 | goto failed_mount_wq; |
eb40a09c JS |
3532 | } |
3533 | ||
d4da6c9c LT |
3534 | if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { |
3535 | jbd2_journal_set_features(sbi->s_journal, | |
3536 | JBD2_FEATURE_COMPAT_CHECKSUM, 0, | |
818d276c | 3537 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); |
d4da6c9c LT |
3538 | } else if (test_opt(sb, JOURNAL_CHECKSUM)) { |
3539 | jbd2_journal_set_features(sbi->s_journal, | |
3540 | JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0); | |
818d276c GS |
3541 | jbd2_journal_clear_features(sbi->s_journal, 0, 0, |
3542 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); | |
d4da6c9c LT |
3543 | } else { |
3544 | jbd2_journal_clear_features(sbi->s_journal, | |
3545 | JBD2_FEATURE_COMPAT_CHECKSUM, 0, | |
3546 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); | |
3547 | } | |
818d276c | 3548 | |
ac27a0ec DK |
3549 | /* We have now updated the journal if required, so we can |
3550 | * validate the data journaling mode. */ | |
3551 | switch (test_opt(sb, DATA_FLAGS)) { | |
3552 | case 0: | |
3553 | /* No mode set, assume a default based on the journal | |
63f57933 AM |
3554 | * capabilities: ORDERED_DATA if the journal can |
3555 | * cope, else JOURNAL_DATA | |
3556 | */ | |
dab291af MC |
3557 | if (jbd2_journal_check_available_features |
3558 | (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) | |
fd8c37ec | 3559 | set_opt(sb, ORDERED_DATA); |
ac27a0ec | 3560 | else |
fd8c37ec | 3561 | set_opt(sb, JOURNAL_DATA); |
ac27a0ec DK |
3562 | break; |
3563 | ||
617ba13b MC |
3564 | case EXT4_MOUNT_ORDERED_DATA: |
3565 | case EXT4_MOUNT_WRITEBACK_DATA: | |
dab291af MC |
3566 | if (!jbd2_journal_check_available_features |
3567 | (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { | |
b31e1552 ES |
3568 | ext4_msg(sb, KERN_ERR, "Journal does not support " |
3569 | "requested data journaling mode"); | |
744692dc | 3570 | goto failed_mount_wq; |
ac27a0ec DK |
3571 | } |
3572 | default: | |
3573 | break; | |
3574 | } | |
b3881f74 | 3575 | set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); |
ac27a0ec | 3576 | |
ce7e010a TT |
3577 | /* |
3578 | * The journal may have updated the bg summary counts, so we | |
3579 | * need to update the global counters. | |
3580 | */ | |
3581 | percpu_counter_set(&sbi->s_freeblocks_counter, | |
3582 | ext4_count_free_blocks(sb)); | |
3583 | percpu_counter_set(&sbi->s_freeinodes_counter, | |
3584 | ext4_count_free_inodes(sb)); | |
3585 | percpu_counter_set(&sbi->s_dirs_counter, | |
3586 | ext4_count_dirs(sb)); | |
3587 | percpu_counter_set(&sbi->s_dirtyblocks_counter, 0); | |
206f7ab4 | 3588 | |
ce7e010a | 3589 | no_journal: |
fd89d5f2 TH |
3590 | /* |
3591 | * The maximum number of concurrent works can be high and | |
3592 | * concurrency isn't really necessary. Limit it to 1. | |
3593 | */ | |
3594 | EXT4_SB(sb)->dio_unwritten_wq = | |
ae005cbe | 3595 | alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); |
4c0425ff MC |
3596 | if (!EXT4_SB(sb)->dio_unwritten_wq) { |
3597 | printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n"); | |
3598 | goto failed_mount_wq; | |
3599 | } | |
3600 | ||
ac27a0ec | 3601 | /* |
dab291af | 3602 | * The jbd2_journal_load will have done any necessary log recovery, |
ac27a0ec DK |
3603 | * so we can safely mount the rest of the filesystem now. |
3604 | */ | |
3605 | ||
1d1fe1ee DH |
3606 | root = ext4_iget(sb, EXT4_ROOT_INO); |
3607 | if (IS_ERR(root)) { | |
b31e1552 | 3608 | ext4_msg(sb, KERN_ERR, "get root inode failed"); |
1d1fe1ee | 3609 | ret = PTR_ERR(root); |
32a9bb57 | 3610 | root = NULL; |
ac27a0ec DK |
3611 | goto failed_mount4; |
3612 | } | |
3613 | if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { | |
b31e1552 | 3614 | ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); |
ac27a0ec DK |
3615 | goto failed_mount4; |
3616 | } | |
1d1fe1ee DH |
3617 | sb->s_root = d_alloc_root(root); |
3618 | if (!sb->s_root) { | |
b31e1552 | 3619 | ext4_msg(sb, KERN_ERR, "get root dentry failed"); |
1d1fe1ee DH |
3620 | ret = -ENOMEM; |
3621 | goto failed_mount4; | |
3622 | } | |
ac27a0ec | 3623 | |
2b2d6d01 | 3624 | ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY); |
ef7f3835 KS |
3625 | |
3626 | /* determine the minimum size of new large inodes, if present */ | |
3627 | if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { | |
3628 | sbi->s_want_extra_isize = sizeof(struct ext4_inode) - | |
3629 | EXT4_GOOD_OLD_INODE_SIZE; | |
3630 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, | |
3631 | EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) { | |
3632 | if (sbi->s_want_extra_isize < | |
3633 | le16_to_cpu(es->s_want_extra_isize)) | |
3634 | sbi->s_want_extra_isize = | |
3635 | le16_to_cpu(es->s_want_extra_isize); | |
3636 | if (sbi->s_want_extra_isize < | |
3637 | le16_to_cpu(es->s_min_extra_isize)) | |
3638 | sbi->s_want_extra_isize = | |
3639 | le16_to_cpu(es->s_min_extra_isize); | |
3640 | } | |
3641 | } | |
3642 | /* Check if enough inode space is available */ | |
3643 | if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize > | |
3644 | sbi->s_inode_size) { | |
3645 | sbi->s_want_extra_isize = sizeof(struct ext4_inode) - | |
3646 | EXT4_GOOD_OLD_INODE_SIZE; | |
b31e1552 ES |
3647 | ext4_msg(sb, KERN_INFO, "required extra inode space not" |
3648 | "available"); | |
ef7f3835 KS |
3649 | } |
3650 | ||
90576c0b TT |
3651 | if (test_opt(sb, DELALLOC) && |
3652 | (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) { | |
b31e1552 ES |
3653 | ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - " |
3654 | "requested data journaling mode"); | |
fd8c37ec | 3655 | clear_opt(sb, DELALLOC); |
90576c0b | 3656 | } |
744692dc JZ |
3657 | if (test_opt(sb, DIOREAD_NOLOCK)) { |
3658 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { | |
3659 | ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock " | |
3660 | "option - requested data journaling mode"); | |
fd8c37ec | 3661 | clear_opt(sb, DIOREAD_NOLOCK); |
744692dc JZ |
3662 | } |
3663 | if (sb->s_blocksize < PAGE_SIZE) { | |
3664 | ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock " | |
3665 | "option - block size is too small"); | |
fd8c37ec | 3666 | clear_opt(sb, DIOREAD_NOLOCK); |
744692dc JZ |
3667 | } |
3668 | } | |
c2774d84 | 3669 | |
6fd058f7 TT |
3670 | err = ext4_setup_system_zone(sb); |
3671 | if (err) { | |
b31e1552 | 3672 | ext4_msg(sb, KERN_ERR, "failed to initialize system " |
fbe845dd | 3673 | "zone (%d)", err); |
6fd058f7 TT |
3674 | goto failed_mount4; |
3675 | } | |
3676 | ||
c2774d84 AK |
3677 | ext4_ext_init(sb); |
3678 | err = ext4_mb_init(sb, needs_recovery); | |
3679 | if (err) { | |
421f91d2 | 3680 | ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", |
b31e1552 | 3681 | err); |
c2774d84 AK |
3682 | goto failed_mount4; |
3683 | } | |
3684 | ||
bfff6873 LC |
3685 | err = ext4_register_li_request(sb, first_not_zeroed); |
3686 | if (err) | |
3687 | goto failed_mount4; | |
3688 | ||
3197ebdb TT |
3689 | sbi->s_kobj.kset = ext4_kset; |
3690 | init_completion(&sbi->s_kobj_unregister); | |
3691 | err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL, | |
3692 | "%s", sb->s_id); | |
3693 | if (err) { | |
3694 | ext4_mb_release(sb); | |
3695 | ext4_ext_release(sb); | |
3696 | goto failed_mount4; | |
3697 | }; | |
3698 | ||
617ba13b MC |
3699 | EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; |
3700 | ext4_orphan_cleanup(sb, es); | |
3701 | EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; | |
0390131b | 3702 | if (needs_recovery) { |
b31e1552 | 3703 | ext4_msg(sb, KERN_INFO, "recovery complete"); |
0390131b FM |
3704 | ext4_mark_recovery_complete(sb, es); |
3705 | } | |
3706 | if (EXT4_SB(sb)->s_journal) { | |
3707 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) | |
3708 | descr = " journalled data mode"; | |
3709 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) | |
3710 | descr = " ordered data mode"; | |
3711 | else | |
3712 | descr = " writeback data mode"; | |
3713 | } else | |
3714 | descr = "out journal"; | |
3715 | ||
d4c402d9 | 3716 | ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " |
8b67f04a TT |
3717 | "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts, |
3718 | *sbi->s_es->s_mount_opts ? "; " : "", orig_data); | |
ac27a0ec | 3719 | |
66e61a9e TT |
3720 | if (es->s_error_count) |
3721 | mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ | |
ac27a0ec | 3722 | |
d4c402d9 | 3723 | kfree(orig_data); |
ac27a0ec DK |
3724 | return 0; |
3725 | ||
617ba13b | 3726 | cantfind_ext4: |
ac27a0ec | 3727 | if (!silent) |
b31e1552 | 3728 | ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); |
ac27a0ec DK |
3729 | goto failed_mount; |
3730 | ||
3731 | failed_mount4: | |
32a9bb57 MK |
3732 | iput(root); |
3733 | sb->s_root = NULL; | |
b31e1552 | 3734 | ext4_msg(sb, KERN_ERR, "mount failed"); |
4c0425ff MC |
3735 | destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq); |
3736 | failed_mount_wq: | |
6fd058f7 | 3737 | ext4_release_system_zone(sb); |
0390131b FM |
3738 | if (sbi->s_journal) { |
3739 | jbd2_journal_destroy(sbi->s_journal); | |
3740 | sbi->s_journal = NULL; | |
3741 | } | |
ac27a0ec | 3742 | failed_mount3: |
04496411 | 3743 | del_timer(&sbi->s_err_report); |
c5ca7c76 TT |
3744 | if (sbi->s_flex_groups) { |
3745 | if (is_vmalloc_addr(sbi->s_flex_groups)) | |
3746 | vfree(sbi->s_flex_groups); | |
3747 | else | |
3748 | kfree(sbi->s_flex_groups); | |
3749 | } | |
ce7e010a TT |
3750 | percpu_counter_destroy(&sbi->s_freeblocks_counter); |
3751 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
3752 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
3753 | percpu_counter_destroy(&sbi->s_dirtyblocks_counter); | |
c5e06d10 JL |
3754 | if (sbi->s_mmp_tsk) |
3755 | kthread_stop(sbi->s_mmp_tsk); | |
ac27a0ec DK |
3756 | failed_mount2: |
3757 | for (i = 0; i < db_count; i++) | |
3758 | brelse(sbi->s_group_desc[i]); | |
3759 | kfree(sbi->s_group_desc); | |
3760 | failed_mount: | |
240799cd | 3761 | if (sbi->s_proc) { |
9f6200bb | 3762 | remove_proc_entry(sb->s_id, ext4_proc_root); |
240799cd | 3763 | } |
ac27a0ec DK |
3764 | #ifdef CONFIG_QUOTA |
3765 | for (i = 0; i < MAXQUOTAS; i++) | |
3766 | kfree(sbi->s_qf_names[i]); | |
3767 | #endif | |
617ba13b | 3768 | ext4_blkdev_remove(sbi); |
ac27a0ec DK |
3769 | brelse(bh); |
3770 | out_fail: | |
3771 | sb->s_fs_info = NULL; | |
f6830165 | 3772 | kfree(sbi->s_blockgroup_lock); |
ac27a0ec | 3773 | kfree(sbi); |
dcc7dae3 | 3774 | out_free_orig: |
d4c402d9 | 3775 | kfree(orig_data); |
1d1fe1ee | 3776 | return ret; |
ac27a0ec DK |
3777 | } |
3778 | ||
3779 | /* | |
3780 | * Setup any per-fs journal parameters now. We'll do this both on | |
3781 | * initial mount, once the journal has been initialised but before we've | |
3782 | * done any recovery; and again on any subsequent remount. | |
3783 | */ | |
617ba13b | 3784 | static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) |
ac27a0ec | 3785 | { |
617ba13b | 3786 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 3787 | |
30773840 TT |
3788 | journal->j_commit_interval = sbi->s_commit_interval; |
3789 | journal->j_min_batch_time = sbi->s_min_batch_time; | |
3790 | journal->j_max_batch_time = sbi->s_max_batch_time; | |
ac27a0ec | 3791 | |
a931da6a | 3792 | write_lock(&journal->j_state_lock); |
ac27a0ec | 3793 | if (test_opt(sb, BARRIER)) |
dab291af | 3794 | journal->j_flags |= JBD2_BARRIER; |
ac27a0ec | 3795 | else |
dab291af | 3796 | journal->j_flags &= ~JBD2_BARRIER; |
5bf5683a HK |
3797 | if (test_opt(sb, DATA_ERR_ABORT)) |
3798 | journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; | |
3799 | else | |
3800 | journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; | |
a931da6a | 3801 | write_unlock(&journal->j_state_lock); |
ac27a0ec DK |
3802 | } |
3803 | ||
617ba13b | 3804 | static journal_t *ext4_get_journal(struct super_block *sb, |
ac27a0ec DK |
3805 | unsigned int journal_inum) |
3806 | { | |
3807 | struct inode *journal_inode; | |
3808 | journal_t *journal; | |
3809 | ||
0390131b FM |
3810 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
3811 | ||
ac27a0ec DK |
3812 | /* First, test for the existence of a valid inode on disk. Bad |
3813 | * things happen if we iget() an unused inode, as the subsequent | |
3814 | * iput() will try to delete it. */ | |
3815 | ||
1d1fe1ee DH |
3816 | journal_inode = ext4_iget(sb, journal_inum); |
3817 | if (IS_ERR(journal_inode)) { | |
b31e1552 | 3818 | ext4_msg(sb, KERN_ERR, "no journal found"); |
ac27a0ec DK |
3819 | return NULL; |
3820 | } | |
3821 | if (!journal_inode->i_nlink) { | |
3822 | make_bad_inode(journal_inode); | |
3823 | iput(journal_inode); | |
b31e1552 | 3824 | ext4_msg(sb, KERN_ERR, "journal inode is deleted"); |
ac27a0ec DK |
3825 | return NULL; |
3826 | } | |
3827 | ||
e5f8eab8 | 3828 | jbd_debug(2, "Journal inode found at %p: %lld bytes\n", |
ac27a0ec | 3829 | journal_inode, journal_inode->i_size); |
1d1fe1ee | 3830 | if (!S_ISREG(journal_inode->i_mode)) { |
b31e1552 | 3831 | ext4_msg(sb, KERN_ERR, "invalid journal inode"); |
ac27a0ec DK |
3832 | iput(journal_inode); |
3833 | return NULL; | |
3834 | } | |
3835 | ||
dab291af | 3836 | journal = jbd2_journal_init_inode(journal_inode); |
ac27a0ec | 3837 | if (!journal) { |
b31e1552 | 3838 | ext4_msg(sb, KERN_ERR, "Could not load journal inode"); |
ac27a0ec DK |
3839 | iput(journal_inode); |
3840 | return NULL; | |
3841 | } | |
3842 | journal->j_private = sb; | |
617ba13b | 3843 | ext4_init_journal_params(sb, journal); |
ac27a0ec DK |
3844 | return journal; |
3845 | } | |
3846 | ||
617ba13b | 3847 | static journal_t *ext4_get_dev_journal(struct super_block *sb, |
ac27a0ec DK |
3848 | dev_t j_dev) |
3849 | { | |
2b2d6d01 | 3850 | struct buffer_head *bh; |
ac27a0ec | 3851 | journal_t *journal; |
617ba13b MC |
3852 | ext4_fsblk_t start; |
3853 | ext4_fsblk_t len; | |
ac27a0ec | 3854 | int hblock, blocksize; |
617ba13b | 3855 | ext4_fsblk_t sb_block; |
ac27a0ec | 3856 | unsigned long offset; |
2b2d6d01 | 3857 | struct ext4_super_block *es; |
ac27a0ec DK |
3858 | struct block_device *bdev; |
3859 | ||
0390131b FM |
3860 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
3861 | ||
b31e1552 | 3862 | bdev = ext4_blkdev_get(j_dev, sb); |
ac27a0ec DK |
3863 | if (bdev == NULL) |
3864 | return NULL; | |
3865 | ||
ac27a0ec | 3866 | blocksize = sb->s_blocksize; |
e1defc4f | 3867 | hblock = bdev_logical_block_size(bdev); |
ac27a0ec | 3868 | if (blocksize < hblock) { |
b31e1552 ES |
3869 | ext4_msg(sb, KERN_ERR, |
3870 | "blocksize too small for journal device"); | |
ac27a0ec DK |
3871 | goto out_bdev; |
3872 | } | |
3873 | ||
617ba13b MC |
3874 | sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; |
3875 | offset = EXT4_MIN_BLOCK_SIZE % blocksize; | |
ac27a0ec DK |
3876 | set_blocksize(bdev, blocksize); |
3877 | if (!(bh = __bread(bdev, sb_block, blocksize))) { | |
b31e1552 ES |
3878 | ext4_msg(sb, KERN_ERR, "couldn't read superblock of " |
3879 | "external journal"); | |
ac27a0ec DK |
3880 | goto out_bdev; |
3881 | } | |
3882 | ||
617ba13b MC |
3883 | es = (struct ext4_super_block *) (((char *)bh->b_data) + offset); |
3884 | if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || | |
ac27a0ec | 3885 | !(le32_to_cpu(es->s_feature_incompat) & |
617ba13b | 3886 | EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { |
b31e1552 ES |
3887 | ext4_msg(sb, KERN_ERR, "external journal has " |
3888 | "bad superblock"); | |
ac27a0ec DK |
3889 | brelse(bh); |
3890 | goto out_bdev; | |
3891 | } | |
3892 | ||
617ba13b | 3893 | if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { |
b31e1552 | 3894 | ext4_msg(sb, KERN_ERR, "journal UUID does not match"); |
ac27a0ec DK |
3895 | brelse(bh); |
3896 | goto out_bdev; | |
3897 | } | |
3898 | ||
bd81d8ee | 3899 | len = ext4_blocks_count(es); |
ac27a0ec DK |
3900 | start = sb_block + 1; |
3901 | brelse(bh); /* we're done with the superblock */ | |
3902 | ||
dab291af | 3903 | journal = jbd2_journal_init_dev(bdev, sb->s_bdev, |
ac27a0ec DK |
3904 | start, len, blocksize); |
3905 | if (!journal) { | |
b31e1552 | 3906 | ext4_msg(sb, KERN_ERR, "failed to create device journal"); |
ac27a0ec DK |
3907 | goto out_bdev; |
3908 | } | |
3909 | journal->j_private = sb; | |
3910 | ll_rw_block(READ, 1, &journal->j_sb_buffer); | |
3911 | wait_on_buffer(journal->j_sb_buffer); | |
3912 | if (!buffer_uptodate(journal->j_sb_buffer)) { | |
b31e1552 | 3913 | ext4_msg(sb, KERN_ERR, "I/O error on journal device"); |
ac27a0ec DK |
3914 | goto out_journal; |
3915 | } | |
3916 | if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { | |
b31e1552 ES |
3917 | ext4_msg(sb, KERN_ERR, "External journal has more than one " |
3918 | "user (unsupported) - %d", | |
ac27a0ec DK |
3919 | be32_to_cpu(journal->j_superblock->s_nr_users)); |
3920 | goto out_journal; | |
3921 | } | |
617ba13b MC |
3922 | EXT4_SB(sb)->journal_bdev = bdev; |
3923 | ext4_init_journal_params(sb, journal); | |
ac27a0ec | 3924 | return journal; |
0b8e58a1 | 3925 | |
ac27a0ec | 3926 | out_journal: |
dab291af | 3927 | jbd2_journal_destroy(journal); |
ac27a0ec | 3928 | out_bdev: |
617ba13b | 3929 | ext4_blkdev_put(bdev); |
ac27a0ec DK |
3930 | return NULL; |
3931 | } | |
3932 | ||
617ba13b MC |
3933 | static int ext4_load_journal(struct super_block *sb, |
3934 | struct ext4_super_block *es, | |
ac27a0ec DK |
3935 | unsigned long journal_devnum) |
3936 | { | |
3937 | journal_t *journal; | |
3938 | unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); | |
3939 | dev_t journal_dev; | |
3940 | int err = 0; | |
3941 | int really_read_only; | |
3942 | ||
0390131b FM |
3943 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
3944 | ||
ac27a0ec DK |
3945 | if (journal_devnum && |
3946 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { | |
b31e1552 ES |
3947 | ext4_msg(sb, KERN_INFO, "external journal device major/minor " |
3948 | "numbers have changed"); | |
ac27a0ec DK |
3949 | journal_dev = new_decode_dev(journal_devnum); |
3950 | } else | |
3951 | journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); | |
3952 | ||
3953 | really_read_only = bdev_read_only(sb->s_bdev); | |
3954 | ||
3955 | /* | |
3956 | * Are we loading a blank journal or performing recovery after a | |
3957 | * crash? For recovery, we need to check in advance whether we | |
3958 | * can get read-write access to the device. | |
3959 | */ | |
617ba13b | 3960 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { |
ac27a0ec | 3961 | if (sb->s_flags & MS_RDONLY) { |
b31e1552 ES |
3962 | ext4_msg(sb, KERN_INFO, "INFO: recovery " |
3963 | "required on readonly filesystem"); | |
ac27a0ec | 3964 | if (really_read_only) { |
b31e1552 ES |
3965 | ext4_msg(sb, KERN_ERR, "write access " |
3966 | "unavailable, cannot proceed"); | |
ac27a0ec DK |
3967 | return -EROFS; |
3968 | } | |
b31e1552 ES |
3969 | ext4_msg(sb, KERN_INFO, "write access will " |
3970 | "be enabled during recovery"); | |
ac27a0ec DK |
3971 | } |
3972 | } | |
3973 | ||
3974 | if (journal_inum && journal_dev) { | |
b31e1552 ES |
3975 | ext4_msg(sb, KERN_ERR, "filesystem has both journal " |
3976 | "and inode journals!"); | |
ac27a0ec DK |
3977 | return -EINVAL; |
3978 | } | |
3979 | ||
3980 | if (journal_inum) { | |
617ba13b | 3981 | if (!(journal = ext4_get_journal(sb, journal_inum))) |
ac27a0ec DK |
3982 | return -EINVAL; |
3983 | } else { | |
617ba13b | 3984 | if (!(journal = ext4_get_dev_journal(sb, journal_dev))) |
ac27a0ec DK |
3985 | return -EINVAL; |
3986 | } | |
3987 | ||
90576c0b | 3988 | if (!(journal->j_flags & JBD2_BARRIER)) |
b31e1552 | 3989 | ext4_msg(sb, KERN_INFO, "barriers disabled"); |
4776004f | 3990 | |
ac27a0ec | 3991 | if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) { |
dab291af | 3992 | err = jbd2_journal_update_format(journal); |
ac27a0ec | 3993 | if (err) { |
b31e1552 | 3994 | ext4_msg(sb, KERN_ERR, "error updating journal"); |
dab291af | 3995 | jbd2_journal_destroy(journal); |
ac27a0ec DK |
3996 | return err; |
3997 | } | |
3998 | } | |
3999 | ||
617ba13b | 4000 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) |
dab291af | 4001 | err = jbd2_journal_wipe(journal, !really_read_only); |
1c13d5c0 TT |
4002 | if (!err) { |
4003 | char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); | |
4004 | if (save) | |
4005 | memcpy(save, ((char *) es) + | |
4006 | EXT4_S_ERR_START, EXT4_S_ERR_LEN); | |
dab291af | 4007 | err = jbd2_journal_load(journal); |
1c13d5c0 TT |
4008 | if (save) |
4009 | memcpy(((char *) es) + EXT4_S_ERR_START, | |
4010 | save, EXT4_S_ERR_LEN); | |
4011 | kfree(save); | |
4012 | } | |
ac27a0ec DK |
4013 | |
4014 | if (err) { | |
b31e1552 | 4015 | ext4_msg(sb, KERN_ERR, "error loading journal"); |
dab291af | 4016 | jbd2_journal_destroy(journal); |
ac27a0ec DK |
4017 | return err; |
4018 | } | |
4019 | ||
617ba13b MC |
4020 | EXT4_SB(sb)->s_journal = journal; |
4021 | ext4_clear_journal_err(sb, es); | |
ac27a0ec | 4022 | |
c41303ce | 4023 | if (!really_read_only && journal_devnum && |
ac27a0ec DK |
4024 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { |
4025 | es->s_journal_dev = cpu_to_le32(journal_devnum); | |
ac27a0ec DK |
4026 | |
4027 | /* Make sure we flush the recovery flag to disk. */ | |
e2d67052 | 4028 | ext4_commit_super(sb, 1); |
ac27a0ec DK |
4029 | } |
4030 | ||
4031 | return 0; | |
4032 | } | |
4033 | ||
e2d67052 | 4034 | static int ext4_commit_super(struct super_block *sb, int sync) |
ac27a0ec | 4035 | { |
e2d67052 | 4036 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
617ba13b | 4037 | struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; |
c4be0c1d | 4038 | int error = 0; |
ac27a0ec DK |
4039 | |
4040 | if (!sbh) | |
c4be0c1d | 4041 | return error; |
914258bf TT |
4042 | if (buffer_write_io_error(sbh)) { |
4043 | /* | |
4044 | * Oh, dear. A previous attempt to write the | |
4045 | * superblock failed. This could happen because the | |
4046 | * USB device was yanked out. Or it could happen to | |
4047 | * be a transient write error and maybe the block will | |
4048 | * be remapped. Nothing we can do but to retry the | |
4049 | * write and hope for the best. | |
4050 | */ | |
b31e1552 ES |
4051 | ext4_msg(sb, KERN_ERR, "previous I/O error to " |
4052 | "superblock detected"); | |
914258bf TT |
4053 | clear_buffer_write_io_error(sbh); |
4054 | set_buffer_uptodate(sbh); | |
4055 | } | |
71290b36 TT |
4056 | /* |
4057 | * If the file system is mounted read-only, don't update the | |
4058 | * superblock write time. This avoids updating the superblock | |
4059 | * write time when we are mounting the root file system | |
4060 | * read/only but we need to replay the journal; at that point, | |
4061 | * for people who are east of GMT and who make their clock | |
4062 | * tick in localtime for Windows bug-for-bug compatibility, | |
4063 | * the clock is set in the future, and this will cause e2fsck | |
4064 | * to complain and force a full file system check. | |
4065 | */ | |
4066 | if (!(sb->s_flags & MS_RDONLY)) | |
4067 | es->s_wtime = cpu_to_le32(get_seconds()); | |
f613dfcb TT |
4068 | if (sb->s_bdev->bd_part) |
4069 | es->s_kbytes_written = | |
4070 | cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + | |
afc32f7e TT |
4071 | ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - |
4072 | EXT4_SB(sb)->s_sectors_written_start) >> 1)); | |
f613dfcb TT |
4073 | else |
4074 | es->s_kbytes_written = | |
4075 | cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); | |
ce7e010a TT |
4076 | ext4_free_blocks_count_set(es, percpu_counter_sum_positive( |
4077 | &EXT4_SB(sb)->s_freeblocks_counter)); | |
4078 | es->s_free_inodes_count = | |
4079 | cpu_to_le32(percpu_counter_sum_positive( | |
4080 | &EXT4_SB(sb)->s_freeinodes_counter)); | |
7234ab2a | 4081 | sb->s_dirt = 0; |
ac27a0ec DK |
4082 | BUFFER_TRACE(sbh, "marking dirty"); |
4083 | mark_buffer_dirty(sbh); | |
914258bf | 4084 | if (sync) { |
c4be0c1d TS |
4085 | error = sync_dirty_buffer(sbh); |
4086 | if (error) | |
4087 | return error; | |
4088 | ||
4089 | error = buffer_write_io_error(sbh); | |
4090 | if (error) { | |
b31e1552 ES |
4091 | ext4_msg(sb, KERN_ERR, "I/O error while writing " |
4092 | "superblock"); | |
914258bf TT |
4093 | clear_buffer_write_io_error(sbh); |
4094 | set_buffer_uptodate(sbh); | |
4095 | } | |
4096 | } | |
c4be0c1d | 4097 | return error; |
ac27a0ec DK |
4098 | } |
4099 | ||
ac27a0ec DK |
4100 | /* |
4101 | * Have we just finished recovery? If so, and if we are mounting (or | |
4102 | * remounting) the filesystem readonly, then we will end up with a | |
4103 | * consistent fs on disk. Record that fact. | |
4104 | */ | |
2b2d6d01 TT |
4105 | static void ext4_mark_recovery_complete(struct super_block *sb, |
4106 | struct ext4_super_block *es) | |
ac27a0ec | 4107 | { |
617ba13b | 4108 | journal_t *journal = EXT4_SB(sb)->s_journal; |
ac27a0ec | 4109 | |
0390131b FM |
4110 | if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { |
4111 | BUG_ON(journal != NULL); | |
4112 | return; | |
4113 | } | |
dab291af | 4114 | jbd2_journal_lock_updates(journal); |
7ffe1ea8 HK |
4115 | if (jbd2_journal_flush(journal) < 0) |
4116 | goto out; | |
4117 | ||
617ba13b | 4118 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) && |
ac27a0ec | 4119 | sb->s_flags & MS_RDONLY) { |
617ba13b | 4120 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); |
e2d67052 | 4121 | ext4_commit_super(sb, 1); |
ac27a0ec | 4122 | } |
7ffe1ea8 HK |
4123 | |
4124 | out: | |
dab291af | 4125 | jbd2_journal_unlock_updates(journal); |
ac27a0ec DK |
4126 | } |
4127 | ||
4128 | /* | |
4129 | * If we are mounting (or read-write remounting) a filesystem whose journal | |
4130 | * has recorded an error from a previous lifetime, move that error to the | |
4131 | * main filesystem now. | |
4132 | */ | |
2b2d6d01 TT |
4133 | static void ext4_clear_journal_err(struct super_block *sb, |
4134 | struct ext4_super_block *es) | |
ac27a0ec DK |
4135 | { |
4136 | journal_t *journal; | |
4137 | int j_errno; | |
4138 | const char *errstr; | |
4139 | ||
0390131b FM |
4140 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
4141 | ||
617ba13b | 4142 | journal = EXT4_SB(sb)->s_journal; |
ac27a0ec DK |
4143 | |
4144 | /* | |
4145 | * Now check for any error status which may have been recorded in the | |
617ba13b | 4146 | * journal by a prior ext4_error() or ext4_abort() |
ac27a0ec DK |
4147 | */ |
4148 | ||
dab291af | 4149 | j_errno = jbd2_journal_errno(journal); |
ac27a0ec DK |
4150 | if (j_errno) { |
4151 | char nbuf[16]; | |
4152 | ||
617ba13b | 4153 | errstr = ext4_decode_error(sb, j_errno, nbuf); |
12062ddd | 4154 | ext4_warning(sb, "Filesystem error recorded " |
ac27a0ec | 4155 | "from previous mount: %s", errstr); |
12062ddd | 4156 | ext4_warning(sb, "Marking fs in need of filesystem check."); |
ac27a0ec | 4157 | |
617ba13b MC |
4158 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; |
4159 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
e2d67052 | 4160 | ext4_commit_super(sb, 1); |
ac27a0ec | 4161 | |
dab291af | 4162 | jbd2_journal_clear_err(journal); |
ac27a0ec DK |
4163 | } |
4164 | } | |
4165 | ||
4166 | /* | |
4167 | * Force the running and committing transactions to commit, | |
4168 | * and wait on the commit. | |
4169 | */ | |
617ba13b | 4170 | int ext4_force_commit(struct super_block *sb) |
ac27a0ec DK |
4171 | { |
4172 | journal_t *journal; | |
0390131b | 4173 | int ret = 0; |
ac27a0ec DK |
4174 | |
4175 | if (sb->s_flags & MS_RDONLY) | |
4176 | return 0; | |
4177 | ||
617ba13b | 4178 | journal = EXT4_SB(sb)->s_journal; |
6b0310fb | 4179 | if (journal) { |
437f88cc | 4180 | vfs_check_frozen(sb, SB_FREEZE_TRANS); |
0390131b | 4181 | ret = ext4_journal_force_commit(journal); |
6b0310fb | 4182 | } |
0390131b | 4183 | |
ac27a0ec DK |
4184 | return ret; |
4185 | } | |
4186 | ||
2b2d6d01 | 4187 | static void ext4_write_super(struct super_block *sb) |
ac27a0ec | 4188 | { |
ebc1ac16 | 4189 | lock_super(sb); |
9ca92389 | 4190 | ext4_commit_super(sb, 1); |
ebc1ac16 | 4191 | unlock_super(sb); |
ac27a0ec DK |
4192 | } |
4193 | ||
617ba13b | 4194 | static int ext4_sync_fs(struct super_block *sb, int wait) |
ac27a0ec | 4195 | { |
14ce0cb4 | 4196 | int ret = 0; |
9eddacf9 | 4197 | tid_t target; |
8d5d02e6 | 4198 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 4199 | |
9bffad1e | 4200 | trace_ext4_sync_fs(sb, wait); |
8d5d02e6 MC |
4201 | flush_workqueue(sbi->dio_unwritten_wq); |
4202 | if (jbd2_journal_start_commit(sbi->s_journal, &target)) { | |
9ca92389 | 4203 | if (wait) |
8d5d02e6 | 4204 | jbd2_log_wait_commit(sbi->s_journal, target); |
0390131b | 4205 | } |
14ce0cb4 | 4206 | return ret; |
ac27a0ec DK |
4207 | } |
4208 | ||
4209 | /* | |
4210 | * LVM calls this function before a (read-only) snapshot is created. This | |
4211 | * gives us a chance to flush the journal completely and mark the fs clean. | |
be4f27d3 YY |
4212 | * |
4213 | * Note that only this function cannot bring a filesystem to be in a clean | |
4214 | * state independently, because ext4 prevents a new handle from being started | |
4215 | * by @sb->s_frozen, which stays in an upper layer. It thus needs help from | |
4216 | * the upper layer. | |
ac27a0ec | 4217 | */ |
c4be0c1d | 4218 | static int ext4_freeze(struct super_block *sb) |
ac27a0ec | 4219 | { |
c4be0c1d TS |
4220 | int error = 0; |
4221 | journal_t *journal; | |
ac27a0ec | 4222 | |
9ca92389 TT |
4223 | if (sb->s_flags & MS_RDONLY) |
4224 | return 0; | |
ac27a0ec | 4225 | |
9ca92389 | 4226 | journal = EXT4_SB(sb)->s_journal; |
7ffe1ea8 | 4227 | |
9ca92389 TT |
4228 | /* Now we set up the journal barrier. */ |
4229 | jbd2_journal_lock_updates(journal); | |
ac27a0ec | 4230 | |
9ca92389 TT |
4231 | /* |
4232 | * Don't clear the needs_recovery flag if we failed to flush | |
4233 | * the journal. | |
4234 | */ | |
4235 | error = jbd2_journal_flush(journal); | |
6b0310fb ES |
4236 | if (error < 0) |
4237 | goto out; | |
9ca92389 TT |
4238 | |
4239 | /* Journal blocked and flushed, clear needs_recovery flag. */ | |
4240 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
4241 | error = ext4_commit_super(sb, 1); | |
6b0310fb ES |
4242 | out: |
4243 | /* we rely on s_frozen to stop further updates */ | |
4244 | jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); | |
4245 | return error; | |
ac27a0ec DK |
4246 | } |
4247 | ||
4248 | /* | |
4249 | * Called by LVM after the snapshot is done. We need to reset the RECOVER | |
4250 | * flag here, even though the filesystem is not technically dirty yet. | |
4251 | */ | |
c4be0c1d | 4252 | static int ext4_unfreeze(struct super_block *sb) |
ac27a0ec | 4253 | { |
9ca92389 TT |
4254 | if (sb->s_flags & MS_RDONLY) |
4255 | return 0; | |
4256 | ||
4257 | lock_super(sb); | |
4258 | /* Reset the needs_recovery flag before the fs is unlocked. */ | |
4259 | EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
4260 | ext4_commit_super(sb, 1); | |
4261 | unlock_super(sb); | |
c4be0c1d | 4262 | return 0; |
ac27a0ec DK |
4263 | } |
4264 | ||
673c6100 TT |
4265 | /* |
4266 | * Structure to save mount options for ext4_remount's benefit | |
4267 | */ | |
4268 | struct ext4_mount_options { | |
4269 | unsigned long s_mount_opt; | |
a2595b8a | 4270 | unsigned long s_mount_opt2; |
673c6100 TT |
4271 | uid_t s_resuid; |
4272 | gid_t s_resgid; | |
4273 | unsigned long s_commit_interval; | |
4274 | u32 s_min_batch_time, s_max_batch_time; | |
4275 | #ifdef CONFIG_QUOTA | |
4276 | int s_jquota_fmt; | |
4277 | char *s_qf_names[MAXQUOTAS]; | |
4278 | #endif | |
4279 | }; | |
4280 | ||
2b2d6d01 | 4281 | static int ext4_remount(struct super_block *sb, int *flags, char *data) |
ac27a0ec | 4282 | { |
2b2d6d01 | 4283 | struct ext4_super_block *es; |
617ba13b MC |
4284 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
4285 | ext4_fsblk_t n_blocks_count = 0; | |
ac27a0ec | 4286 | unsigned long old_sb_flags; |
617ba13b | 4287 | struct ext4_mount_options old_opts; |
c79d967d | 4288 | int enable_quota = 0; |
8a266467 | 4289 | ext4_group_t g; |
b3881f74 | 4290 | unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; |
c5e06d10 | 4291 | int err = 0; |
ac27a0ec DK |
4292 | #ifdef CONFIG_QUOTA |
4293 | int i; | |
4294 | #endif | |
d4c402d9 | 4295 | char *orig_data = kstrdup(data, GFP_KERNEL); |
ac27a0ec DK |
4296 | |
4297 | /* Store the original options */ | |
bbd6851a | 4298 | lock_super(sb); |
ac27a0ec DK |
4299 | old_sb_flags = sb->s_flags; |
4300 | old_opts.s_mount_opt = sbi->s_mount_opt; | |
a2595b8a | 4301 | old_opts.s_mount_opt2 = sbi->s_mount_opt2; |
ac27a0ec DK |
4302 | old_opts.s_resuid = sbi->s_resuid; |
4303 | old_opts.s_resgid = sbi->s_resgid; | |
4304 | old_opts.s_commit_interval = sbi->s_commit_interval; | |
30773840 TT |
4305 | old_opts.s_min_batch_time = sbi->s_min_batch_time; |
4306 | old_opts.s_max_batch_time = sbi->s_max_batch_time; | |
ac27a0ec DK |
4307 | #ifdef CONFIG_QUOTA |
4308 | old_opts.s_jquota_fmt = sbi->s_jquota_fmt; | |
4309 | for (i = 0; i < MAXQUOTAS; i++) | |
4310 | old_opts.s_qf_names[i] = sbi->s_qf_names[i]; | |
4311 | #endif | |
b3881f74 TT |
4312 | if (sbi->s_journal && sbi->s_journal->j_task->io_context) |
4313 | journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; | |
ac27a0ec DK |
4314 | |
4315 | /* | |
4316 | * Allow the "check" option to be passed as a remount option. | |
4317 | */ | |
b3881f74 TT |
4318 | if (!parse_options(data, sb, NULL, &journal_ioprio, |
4319 | &n_blocks_count, 1)) { | |
ac27a0ec DK |
4320 | err = -EINVAL; |
4321 | goto restore_opts; | |
4322 | } | |
4323 | ||
4ab2f15b | 4324 | if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) |
c67d859e | 4325 | ext4_abort(sb, "Abort forced by user"); |
ac27a0ec DK |
4326 | |
4327 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
482a7425 | 4328 | (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); |
ac27a0ec DK |
4329 | |
4330 | es = sbi->s_es; | |
4331 | ||
b3881f74 | 4332 | if (sbi->s_journal) { |
0390131b | 4333 | ext4_init_journal_params(sb, sbi->s_journal); |
b3881f74 TT |
4334 | set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); |
4335 | } | |
ac27a0ec DK |
4336 | |
4337 | if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) || | |
bd81d8ee | 4338 | n_blocks_count > ext4_blocks_count(es)) { |
4ab2f15b | 4339 | if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { |
ac27a0ec DK |
4340 | err = -EROFS; |
4341 | goto restore_opts; | |
4342 | } | |
4343 | ||
4344 | if (*flags & MS_RDONLY) { | |
0f0dd62f CH |
4345 | err = dquot_suspend(sb, -1); |
4346 | if (err < 0) | |
c79d967d | 4347 | goto restore_opts; |
c79d967d | 4348 | |
ac27a0ec DK |
4349 | /* |
4350 | * First of all, the unconditional stuff we have to do | |
4351 | * to disable replay of the journal when we next remount | |
4352 | */ | |
4353 | sb->s_flags |= MS_RDONLY; | |
4354 | ||
4355 | /* | |
4356 | * OK, test if we are remounting a valid rw partition | |
4357 | * readonly, and if so set the rdonly flag and then | |
4358 | * mark the partition as valid again. | |
4359 | */ | |
617ba13b MC |
4360 | if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && |
4361 | (sbi->s_mount_state & EXT4_VALID_FS)) | |
ac27a0ec DK |
4362 | es->s_state = cpu_to_le16(sbi->s_mount_state); |
4363 | ||
a63c9eb2 | 4364 | if (sbi->s_journal) |
0390131b | 4365 | ext4_mark_recovery_complete(sb, es); |
ac27a0ec | 4366 | } else { |
a13fb1a4 ES |
4367 | /* Make sure we can mount this feature set readwrite */ |
4368 | if (!ext4_feature_set_ok(sb, 0)) { | |
ac27a0ec DK |
4369 | err = -EROFS; |
4370 | goto restore_opts; | |
4371 | } | |
8a266467 TT |
4372 | /* |
4373 | * Make sure the group descriptor checksums | |
0b8e58a1 | 4374 | * are sane. If they aren't, refuse to remount r/w. |
8a266467 TT |
4375 | */ |
4376 | for (g = 0; g < sbi->s_groups_count; g++) { | |
4377 | struct ext4_group_desc *gdp = | |
4378 | ext4_get_group_desc(sb, g, NULL); | |
4379 | ||
4380 | if (!ext4_group_desc_csum_verify(sbi, g, gdp)) { | |
b31e1552 ES |
4381 | ext4_msg(sb, KERN_ERR, |
4382 | "ext4_remount: Checksum for group %u failed (%u!=%u)", | |
8a266467 TT |
4383 | g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)), |
4384 | le16_to_cpu(gdp->bg_checksum)); | |
4385 | err = -EINVAL; | |
4386 | goto restore_opts; | |
4387 | } | |
4388 | } | |
4389 | ||
ead6596b ES |
4390 | /* |
4391 | * If we have an unprocessed orphan list hanging | |
4392 | * around from a previously readonly bdev mount, | |
4393 | * require a full umount/remount for now. | |
4394 | */ | |
4395 | if (es->s_last_orphan) { | |
b31e1552 | 4396 | ext4_msg(sb, KERN_WARNING, "Couldn't " |
ead6596b ES |
4397 | "remount RDWR because of unprocessed " |
4398 | "orphan inode list. Please " | |
b31e1552 | 4399 | "umount/remount instead"); |
ead6596b ES |
4400 | err = -EINVAL; |
4401 | goto restore_opts; | |
4402 | } | |
4403 | ||
ac27a0ec DK |
4404 | /* |
4405 | * Mounting a RDONLY partition read-write, so reread | |
4406 | * and store the current valid flag. (It may have | |
4407 | * been changed by e2fsck since we originally mounted | |
4408 | * the partition.) | |
4409 | */ | |
0390131b FM |
4410 | if (sbi->s_journal) |
4411 | ext4_clear_journal_err(sb, es); | |
ac27a0ec | 4412 | sbi->s_mount_state = le16_to_cpu(es->s_state); |
617ba13b | 4413 | if ((err = ext4_group_extend(sb, es, n_blocks_count))) |
ac27a0ec | 4414 | goto restore_opts; |
2b2d6d01 | 4415 | if (!ext4_setup_super(sb, es, 0)) |
ac27a0ec | 4416 | sb->s_flags &= ~MS_RDONLY; |
c5e06d10 JL |
4417 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, |
4418 | EXT4_FEATURE_INCOMPAT_MMP)) | |
4419 | if (ext4_multi_mount_protect(sb, | |
4420 | le64_to_cpu(es->s_mmp_block))) { | |
4421 | err = -EROFS; | |
4422 | goto restore_opts; | |
4423 | } | |
c79d967d | 4424 | enable_quota = 1; |
ac27a0ec DK |
4425 | } |
4426 | } | |
bfff6873 LC |
4427 | |
4428 | /* | |
4429 | * Reinitialize lazy itable initialization thread based on | |
4430 | * current settings | |
4431 | */ | |
4432 | if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE)) | |
4433 | ext4_unregister_li_request(sb); | |
4434 | else { | |
4435 | ext4_group_t first_not_zeroed; | |
4436 | first_not_zeroed = ext4_has_uninit_itable(sb); | |
4437 | ext4_register_li_request(sb, first_not_zeroed); | |
4438 | } | |
4439 | ||
6fd058f7 | 4440 | ext4_setup_system_zone(sb); |
0390131b | 4441 | if (sbi->s_journal == NULL) |
e2d67052 | 4442 | ext4_commit_super(sb, 1); |
0390131b | 4443 | |
ac27a0ec DK |
4444 | #ifdef CONFIG_QUOTA |
4445 | /* Release old quota file names */ | |
4446 | for (i = 0; i < MAXQUOTAS; i++) | |
4447 | if (old_opts.s_qf_names[i] && | |
4448 | old_opts.s_qf_names[i] != sbi->s_qf_names[i]) | |
4449 | kfree(old_opts.s_qf_names[i]); | |
4450 | #endif | |
bbd6851a | 4451 | unlock_super(sb); |
c79d967d | 4452 | if (enable_quota) |
0f0dd62f | 4453 | dquot_resume(sb, -1); |
d4c402d9 CW |
4454 | |
4455 | ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); | |
4456 | kfree(orig_data); | |
ac27a0ec | 4457 | return 0; |
0b8e58a1 | 4458 | |
ac27a0ec DK |
4459 | restore_opts: |
4460 | sb->s_flags = old_sb_flags; | |
4461 | sbi->s_mount_opt = old_opts.s_mount_opt; | |
a2595b8a | 4462 | sbi->s_mount_opt2 = old_opts.s_mount_opt2; |
ac27a0ec DK |
4463 | sbi->s_resuid = old_opts.s_resuid; |
4464 | sbi->s_resgid = old_opts.s_resgid; | |
4465 | sbi->s_commit_interval = old_opts.s_commit_interval; | |
30773840 TT |
4466 | sbi->s_min_batch_time = old_opts.s_min_batch_time; |
4467 | sbi->s_max_batch_time = old_opts.s_max_batch_time; | |
ac27a0ec DK |
4468 | #ifdef CONFIG_QUOTA |
4469 | sbi->s_jquota_fmt = old_opts.s_jquota_fmt; | |
4470 | for (i = 0; i < MAXQUOTAS; i++) { | |
4471 | if (sbi->s_qf_names[i] && | |
4472 | old_opts.s_qf_names[i] != sbi->s_qf_names[i]) | |
4473 | kfree(sbi->s_qf_names[i]); | |
4474 | sbi->s_qf_names[i] = old_opts.s_qf_names[i]; | |
4475 | } | |
4476 | #endif | |
bbd6851a | 4477 | unlock_super(sb); |
d4c402d9 | 4478 | kfree(orig_data); |
ac27a0ec DK |
4479 | return err; |
4480 | } | |
4481 | ||
2b2d6d01 | 4482 | static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) |
ac27a0ec DK |
4483 | { |
4484 | struct super_block *sb = dentry->d_sb; | |
617ba13b MC |
4485 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
4486 | struct ext4_super_block *es = sbi->s_es; | |
960cc398 | 4487 | u64 fsid; |
d02a9391 | 4488 | s64 bfree; |
ac27a0ec | 4489 | |
5e70030d BP |
4490 | if (test_opt(sb, MINIX_DF)) { |
4491 | sbi->s_overhead_last = 0; | |
6bc9feff | 4492 | } else if (sbi->s_blocks_last != ext4_blocks_count(es)) { |
8df9675f | 4493 | ext4_group_t i, ngroups = ext4_get_groups_count(sb); |
5e70030d | 4494 | ext4_fsblk_t overhead = 0; |
ac27a0ec DK |
4495 | |
4496 | /* | |
5e70030d BP |
4497 | * Compute the overhead (FS structures). This is constant |
4498 | * for a given filesystem unless the number of block groups | |
4499 | * changes so we cache the previous value until it does. | |
ac27a0ec DK |
4500 | */ |
4501 | ||
4502 | /* | |
4503 | * All of the blocks before first_data_block are | |
4504 | * overhead | |
4505 | */ | |
4506 | overhead = le32_to_cpu(es->s_first_data_block); | |
4507 | ||
4508 | /* | |
4509 | * Add the overhead attributed to the superblock and | |
4510 | * block group descriptors. If the sparse superblocks | |
4511 | * feature is turned on, then not all groups have this. | |
4512 | */ | |
4513 | for (i = 0; i < ngroups; i++) { | |
617ba13b MC |
4514 | overhead += ext4_bg_has_super(sb, i) + |
4515 | ext4_bg_num_gdb(sb, i); | |
ac27a0ec DK |
4516 | cond_resched(); |
4517 | } | |
4518 | ||
4519 | /* | |
4520 | * Every block group has an inode bitmap, a block | |
4521 | * bitmap, and an inode table. | |
4522 | */ | |
5e70030d BP |
4523 | overhead += ngroups * (2 + sbi->s_itb_per_group); |
4524 | sbi->s_overhead_last = overhead; | |
4525 | smp_wmb(); | |
6bc9feff | 4526 | sbi->s_blocks_last = ext4_blocks_count(es); |
ac27a0ec DK |
4527 | } |
4528 | ||
617ba13b | 4529 | buf->f_type = EXT4_SUPER_MAGIC; |
ac27a0ec | 4530 | buf->f_bsize = sb->s_blocksize; |
5e70030d | 4531 | buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last; |
d02a9391 | 4532 | bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) - |
6bc6e63f | 4533 | percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter); |
d02a9391 KM |
4534 | /* prevent underflow in case that few free space is available */ |
4535 | buf->f_bfree = max_t(s64, bfree, 0); | |
bd81d8ee LV |
4536 | buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es); |
4537 | if (buf->f_bfree < ext4_r_blocks_count(es)) | |
ac27a0ec DK |
4538 | buf->f_bavail = 0; |
4539 | buf->f_files = le32_to_cpu(es->s_inodes_count); | |
52d9f3b4 | 4540 | buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); |
617ba13b | 4541 | buf->f_namelen = EXT4_NAME_LEN; |
960cc398 PE |
4542 | fsid = le64_to_cpup((void *)es->s_uuid) ^ |
4543 | le64_to_cpup((void *)es->s_uuid + sizeof(u64)); | |
4544 | buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; | |
4545 | buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; | |
0b8e58a1 | 4546 | |
ac27a0ec DK |
4547 | return 0; |
4548 | } | |
4549 | ||
0b8e58a1 AD |
4550 | /* Helper function for writing quotas on sync - we need to start transaction |
4551 | * before quota file is locked for write. Otherwise the are possible deadlocks: | |
ac27a0ec | 4552 | * Process 1 Process 2 |
617ba13b | 4553 | * ext4_create() quota_sync() |
a269eb18 | 4554 | * jbd2_journal_start() write_dquot() |
871a2931 | 4555 | * dquot_initialize() down(dqio_mutex) |
dab291af | 4556 | * down(dqio_mutex) jbd2_journal_start() |
ac27a0ec DK |
4557 | * |
4558 | */ | |
4559 | ||
4560 | #ifdef CONFIG_QUOTA | |
4561 | ||
4562 | static inline struct inode *dquot_to_inode(struct dquot *dquot) | |
4563 | { | |
4564 | return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type]; | |
4565 | } | |
4566 | ||
617ba13b | 4567 | static int ext4_write_dquot(struct dquot *dquot) |
ac27a0ec DK |
4568 | { |
4569 | int ret, err; | |
4570 | handle_t *handle; | |
4571 | struct inode *inode; | |
4572 | ||
4573 | inode = dquot_to_inode(dquot); | |
617ba13b | 4574 | handle = ext4_journal_start(inode, |
0b8e58a1 | 4575 | EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); |
ac27a0ec DK |
4576 | if (IS_ERR(handle)) |
4577 | return PTR_ERR(handle); | |
4578 | ret = dquot_commit(dquot); | |
617ba13b | 4579 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
4580 | if (!ret) |
4581 | ret = err; | |
4582 | return ret; | |
4583 | } | |
4584 | ||
617ba13b | 4585 | static int ext4_acquire_dquot(struct dquot *dquot) |
ac27a0ec DK |
4586 | { |
4587 | int ret, err; | |
4588 | handle_t *handle; | |
4589 | ||
617ba13b | 4590 | handle = ext4_journal_start(dquot_to_inode(dquot), |
0b8e58a1 | 4591 | EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); |
ac27a0ec DK |
4592 | if (IS_ERR(handle)) |
4593 | return PTR_ERR(handle); | |
4594 | ret = dquot_acquire(dquot); | |
617ba13b | 4595 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
4596 | if (!ret) |
4597 | ret = err; | |
4598 | return ret; | |
4599 | } | |
4600 | ||
617ba13b | 4601 | static int ext4_release_dquot(struct dquot *dquot) |
ac27a0ec DK |
4602 | { |
4603 | int ret, err; | |
4604 | handle_t *handle; | |
4605 | ||
617ba13b | 4606 | handle = ext4_journal_start(dquot_to_inode(dquot), |
0b8e58a1 | 4607 | EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); |
9c3013e9 JK |
4608 | if (IS_ERR(handle)) { |
4609 | /* Release dquot anyway to avoid endless cycle in dqput() */ | |
4610 | dquot_release(dquot); | |
ac27a0ec | 4611 | return PTR_ERR(handle); |
9c3013e9 | 4612 | } |
ac27a0ec | 4613 | ret = dquot_release(dquot); |
617ba13b | 4614 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
4615 | if (!ret) |
4616 | ret = err; | |
4617 | return ret; | |
4618 | } | |
4619 | ||
617ba13b | 4620 | static int ext4_mark_dquot_dirty(struct dquot *dquot) |
ac27a0ec | 4621 | { |
2c8be6b2 | 4622 | /* Are we journaling quotas? */ |
617ba13b MC |
4623 | if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] || |
4624 | EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) { | |
ac27a0ec | 4625 | dquot_mark_dquot_dirty(dquot); |
617ba13b | 4626 | return ext4_write_dquot(dquot); |
ac27a0ec DK |
4627 | } else { |
4628 | return dquot_mark_dquot_dirty(dquot); | |
4629 | } | |
4630 | } | |
4631 | ||
617ba13b | 4632 | static int ext4_write_info(struct super_block *sb, int type) |
ac27a0ec DK |
4633 | { |
4634 | int ret, err; | |
4635 | handle_t *handle; | |
4636 | ||
4637 | /* Data block + inode block */ | |
617ba13b | 4638 | handle = ext4_journal_start(sb->s_root->d_inode, 2); |
ac27a0ec DK |
4639 | if (IS_ERR(handle)) |
4640 | return PTR_ERR(handle); | |
4641 | ret = dquot_commit_info(sb, type); | |
617ba13b | 4642 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
4643 | if (!ret) |
4644 | ret = err; | |
4645 | return ret; | |
4646 | } | |
4647 | ||
4648 | /* | |
4649 | * Turn on quotas during mount time - we need to find | |
4650 | * the quota file and such... | |
4651 | */ | |
617ba13b | 4652 | static int ext4_quota_on_mount(struct super_block *sb, int type) |
ac27a0ec | 4653 | { |
287a8095 CH |
4654 | return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type], |
4655 | EXT4_SB(sb)->s_jquota_fmt, type); | |
ac27a0ec DK |
4656 | } |
4657 | ||
4658 | /* | |
4659 | * Standard function to be called on quota_on | |
4660 | */ | |
617ba13b | 4661 | static int ext4_quota_on(struct super_block *sb, int type, int format_id, |
f00c9e44 | 4662 | struct path *path) |
ac27a0ec DK |
4663 | { |
4664 | int err; | |
ac27a0ec DK |
4665 | |
4666 | if (!test_opt(sb, QUOTA)) | |
4667 | return -EINVAL; | |
0623543b | 4668 | |
ac27a0ec | 4669 | /* Quotafile not on the same filesystem? */ |
f00c9e44 | 4670 | if (path->mnt->mnt_sb != sb) |
ac27a0ec | 4671 | return -EXDEV; |
0623543b JK |
4672 | /* Journaling quota? */ |
4673 | if (EXT4_SB(sb)->s_qf_names[type]) { | |
2b2d6d01 | 4674 | /* Quotafile not in fs root? */ |
f00c9e44 | 4675 | if (path->dentry->d_parent != sb->s_root) |
b31e1552 ES |
4676 | ext4_msg(sb, KERN_WARNING, |
4677 | "Quota file not on filesystem root. " | |
4678 | "Journaled quota will not work"); | |
2b2d6d01 | 4679 | } |
0623543b JK |
4680 | |
4681 | /* | |
4682 | * When we journal data on quota file, we have to flush journal to see | |
4683 | * all updates to the file when we bypass pagecache... | |
4684 | */ | |
0390131b | 4685 | if (EXT4_SB(sb)->s_journal && |
f00c9e44 | 4686 | ext4_should_journal_data(path->dentry->d_inode)) { |
0623543b JK |
4687 | /* |
4688 | * We don't need to lock updates but journal_flush() could | |
4689 | * otherwise be livelocked... | |
4690 | */ | |
4691 | jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); | |
7ffe1ea8 | 4692 | err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); |
0623543b | 4693 | jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); |
f00c9e44 | 4694 | if (err) |
7ffe1ea8 | 4695 | return err; |
0623543b JK |
4696 | } |
4697 | ||
f00c9e44 | 4698 | return dquot_quota_on(sb, type, format_id, path); |
ac27a0ec DK |
4699 | } |
4700 | ||
ca0e05e4 DM |
4701 | static int ext4_quota_off(struct super_block *sb, int type) |
4702 | { | |
21f97697 JK |
4703 | struct inode *inode = sb_dqopt(sb)->files[type]; |
4704 | handle_t *handle; | |
4705 | ||
87009d86 DM |
4706 | /* Force all delayed allocation blocks to be allocated. |
4707 | * Caller already holds s_umount sem */ | |
4708 | if (test_opt(sb, DELALLOC)) | |
ca0e05e4 | 4709 | sync_filesystem(sb); |
ca0e05e4 | 4710 | |
0b268590 AG |
4711 | if (!inode) |
4712 | goto out; | |
4713 | ||
21f97697 JK |
4714 | /* Update modification times of quota files when userspace can |
4715 | * start looking at them */ | |
4716 | handle = ext4_journal_start(inode, 1); | |
4717 | if (IS_ERR(handle)) | |
4718 | goto out; | |
4719 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; | |
4720 | ext4_mark_inode_dirty(handle, inode); | |
4721 | ext4_journal_stop(handle); | |
4722 | ||
4723 | out: | |
ca0e05e4 DM |
4724 | return dquot_quota_off(sb, type); |
4725 | } | |
4726 | ||
ac27a0ec DK |
4727 | /* Read data from quotafile - avoid pagecache and such because we cannot afford |
4728 | * acquiring the locks... As quota files are never truncated and quota code | |
25985edc | 4729 | * itself serializes the operations (and no one else should touch the files) |
ac27a0ec | 4730 | * we don't have to be afraid of races */ |
617ba13b | 4731 | static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, |
ac27a0ec DK |
4732 | size_t len, loff_t off) |
4733 | { | |
4734 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
725d26d3 | 4735 | ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); |
ac27a0ec DK |
4736 | int err = 0; |
4737 | int offset = off & (sb->s_blocksize - 1); | |
4738 | int tocopy; | |
4739 | size_t toread; | |
4740 | struct buffer_head *bh; | |
4741 | loff_t i_size = i_size_read(inode); | |
4742 | ||
4743 | if (off > i_size) | |
4744 | return 0; | |
4745 | if (off+len > i_size) | |
4746 | len = i_size-off; | |
4747 | toread = len; | |
4748 | while (toread > 0) { | |
4749 | tocopy = sb->s_blocksize - offset < toread ? | |
4750 | sb->s_blocksize - offset : toread; | |
617ba13b | 4751 | bh = ext4_bread(NULL, inode, blk, 0, &err); |
ac27a0ec DK |
4752 | if (err) |
4753 | return err; | |
4754 | if (!bh) /* A hole? */ | |
4755 | memset(data, 0, tocopy); | |
4756 | else | |
4757 | memcpy(data, bh->b_data+offset, tocopy); | |
4758 | brelse(bh); | |
4759 | offset = 0; | |
4760 | toread -= tocopy; | |
4761 | data += tocopy; | |
4762 | blk++; | |
4763 | } | |
4764 | return len; | |
4765 | } | |
4766 | ||
4767 | /* Write to quotafile (we know the transaction is already started and has | |
4768 | * enough credits) */ | |
617ba13b | 4769 | static ssize_t ext4_quota_write(struct super_block *sb, int type, |
ac27a0ec DK |
4770 | const char *data, size_t len, loff_t off) |
4771 | { | |
4772 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
725d26d3 | 4773 | ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); |
ac27a0ec DK |
4774 | int err = 0; |
4775 | int offset = off & (sb->s_blocksize - 1); | |
ac27a0ec DK |
4776 | struct buffer_head *bh; |
4777 | handle_t *handle = journal_current_handle(); | |
4778 | ||
0390131b | 4779 | if (EXT4_SB(sb)->s_journal && !handle) { |
b31e1552 ES |
4780 | ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" |
4781 | " cancelled because transaction is not started", | |
9c3013e9 JK |
4782 | (unsigned long long)off, (unsigned long long)len); |
4783 | return -EIO; | |
4784 | } | |
67eeb568 DM |
4785 | /* |
4786 | * Since we account only one data block in transaction credits, | |
4787 | * then it is impossible to cross a block boundary. | |
4788 | */ | |
4789 | if (sb->s_blocksize - offset < len) { | |
4790 | ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" | |
4791 | " cancelled because not block aligned", | |
4792 | (unsigned long long)off, (unsigned long long)len); | |
4793 | return -EIO; | |
4794 | } | |
4795 | ||
ac27a0ec | 4796 | mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA); |
67eeb568 DM |
4797 | bh = ext4_bread(handle, inode, blk, 1, &err); |
4798 | if (!bh) | |
4799 | goto out; | |
62d2b5f2 JK |
4800 | err = ext4_journal_get_write_access(handle, bh); |
4801 | if (err) { | |
4802 | brelse(bh); | |
4803 | goto out; | |
ac27a0ec | 4804 | } |
67eeb568 DM |
4805 | lock_buffer(bh); |
4806 | memcpy(bh->b_data+offset, data, len); | |
4807 | flush_dcache_page(bh->b_page); | |
4808 | unlock_buffer(bh); | |
62d2b5f2 | 4809 | err = ext4_handle_dirty_metadata(handle, NULL, bh); |
67eeb568 | 4810 | brelse(bh); |
ac27a0ec | 4811 | out: |
67eeb568 | 4812 | if (err) { |
4d04e4fb | 4813 | mutex_unlock(&inode->i_mutex); |
ac27a0ec | 4814 | return err; |
4d04e4fb | 4815 | } |
67eeb568 DM |
4816 | if (inode->i_size < off + len) { |
4817 | i_size_write(inode, off + len); | |
617ba13b | 4818 | EXT4_I(inode)->i_disksize = inode->i_size; |
21f97697 | 4819 | ext4_mark_inode_dirty(handle, inode); |
ac27a0ec | 4820 | } |
ac27a0ec | 4821 | mutex_unlock(&inode->i_mutex); |
67eeb568 | 4822 | return len; |
ac27a0ec DK |
4823 | } |
4824 | ||
4825 | #endif | |
4826 | ||
152a0836 AV |
4827 | static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, |
4828 | const char *dev_name, void *data) | |
ac27a0ec | 4829 | { |
152a0836 | 4830 | return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); |
ac27a0ec DK |
4831 | } |
4832 | ||
37f328eb | 4833 | #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) |
24b58424 TT |
4834 | static inline void register_as_ext2(void) |
4835 | { | |
4836 | int err = register_filesystem(&ext2_fs_type); | |
4837 | if (err) | |
4838 | printk(KERN_WARNING | |
4839 | "EXT4-fs: Unable to register as ext2 (%d)\n", err); | |
4840 | } | |
4841 | ||
4842 | static inline void unregister_as_ext2(void) | |
4843 | { | |
4844 | unregister_filesystem(&ext2_fs_type); | |
4845 | } | |
2035e776 TT |
4846 | |
4847 | static inline int ext2_feature_set_ok(struct super_block *sb) | |
4848 | { | |
4849 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP)) | |
4850 | return 0; | |
4851 | if (sb->s_flags & MS_RDONLY) | |
4852 | return 1; | |
4853 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP)) | |
4854 | return 0; | |
4855 | return 1; | |
4856 | } | |
51b7e3c9 | 4857 | MODULE_ALIAS("ext2"); |
24b58424 TT |
4858 | #else |
4859 | static inline void register_as_ext2(void) { } | |
4860 | static inline void unregister_as_ext2(void) { } | |
2035e776 | 4861 | static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } |
24b58424 TT |
4862 | #endif |
4863 | ||
37f328eb | 4864 | #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) |
24b58424 TT |
4865 | static inline void register_as_ext3(void) |
4866 | { | |
4867 | int err = register_filesystem(&ext3_fs_type); | |
4868 | if (err) | |
4869 | printk(KERN_WARNING | |
4870 | "EXT4-fs: Unable to register as ext3 (%d)\n", err); | |
4871 | } | |
4872 | ||
4873 | static inline void unregister_as_ext3(void) | |
4874 | { | |
4875 | unregister_filesystem(&ext3_fs_type); | |
4876 | } | |
2035e776 TT |
4877 | |
4878 | static inline int ext3_feature_set_ok(struct super_block *sb) | |
4879 | { | |
4880 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP)) | |
4881 | return 0; | |
4882 | if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) | |
4883 | return 0; | |
4884 | if (sb->s_flags & MS_RDONLY) | |
4885 | return 1; | |
4886 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP)) | |
4887 | return 0; | |
4888 | return 1; | |
4889 | } | |
51b7e3c9 | 4890 | MODULE_ALIAS("ext3"); |
24b58424 TT |
4891 | #else |
4892 | static inline void register_as_ext3(void) { } | |
4893 | static inline void unregister_as_ext3(void) { } | |
2035e776 | 4894 | static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; } |
24b58424 TT |
4895 | #endif |
4896 | ||
03010a33 TT |
4897 | static struct file_system_type ext4_fs_type = { |
4898 | .owner = THIS_MODULE, | |
4899 | .name = "ext4", | |
152a0836 | 4900 | .mount = ext4_mount, |
03010a33 TT |
4901 | .kill_sb = kill_block_super, |
4902 | .fs_flags = FS_REQUIRES_DEV, | |
4903 | }; | |
4904 | ||
8f021222 | 4905 | static int __init ext4_init_feat_adverts(void) |
857ac889 LC |
4906 | { |
4907 | struct ext4_features *ef; | |
4908 | int ret = -ENOMEM; | |
4909 | ||
4910 | ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL); | |
4911 | if (!ef) | |
4912 | goto out; | |
4913 | ||
4914 | ef->f_kobj.kset = ext4_kset; | |
4915 | init_completion(&ef->f_kobj_unregister); | |
4916 | ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL, | |
4917 | "features"); | |
4918 | if (ret) { | |
4919 | kfree(ef); | |
4920 | goto out; | |
4921 | } | |
4922 | ||
4923 | ext4_feat = ef; | |
4924 | ret = 0; | |
4925 | out: | |
4926 | return ret; | |
4927 | } | |
4928 | ||
8f021222 LC |
4929 | static void ext4_exit_feat_adverts(void) |
4930 | { | |
4931 | kobject_put(&ext4_feat->f_kobj); | |
4932 | wait_for_completion(&ext4_feat->f_kobj_unregister); | |
4933 | kfree(ext4_feat); | |
4934 | } | |
4935 | ||
e9e3bcec ES |
4936 | /* Shared across all ext4 file systems */ |
4937 | wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; | |
4938 | struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ]; | |
4939 | ||
5dabfc78 | 4940 | static int __init ext4_init_fs(void) |
ac27a0ec | 4941 | { |
e9e3bcec | 4942 | int i, err; |
c9de560d | 4943 | |
12e9b892 | 4944 | ext4_check_flag_values(); |
e9e3bcec ES |
4945 | |
4946 | for (i = 0; i < EXT4_WQ_HASH_SZ; i++) { | |
4947 | mutex_init(&ext4__aio_mutex[i]); | |
4948 | init_waitqueue_head(&ext4__ioend_wq[i]); | |
4949 | } | |
4950 | ||
5dabfc78 | 4951 | err = ext4_init_pageio(); |
6fd058f7 TT |
4952 | if (err) |
4953 | return err; | |
5dabfc78 | 4954 | err = ext4_init_system_zone(); |
bd2d0210 | 4955 | if (err) |
dd68314c | 4956 | goto out7; |
3197ebdb TT |
4957 | ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj); |
4958 | if (!ext4_kset) | |
dd68314c | 4959 | goto out6; |
9f6200bb | 4960 | ext4_proc_root = proc_mkdir("fs/ext4", NULL); |
dd68314c TT |
4961 | if (!ext4_proc_root) |
4962 | goto out5; | |
857ac889 LC |
4963 | |
4964 | err = ext4_init_feat_adverts(); | |
dd68314c TT |
4965 | if (err) |
4966 | goto out4; | |
857ac889 | 4967 | |
5dabfc78 | 4968 | err = ext4_init_mballoc(); |
ac27a0ec | 4969 | if (err) |
6fd058f7 | 4970 | goto out3; |
c9de560d | 4971 | |
5dabfc78 | 4972 | err = ext4_init_xattr(); |
c9de560d AT |
4973 | if (err) |
4974 | goto out2; | |
ac27a0ec DK |
4975 | err = init_inodecache(); |
4976 | if (err) | |
4977 | goto out1; | |
24b58424 | 4978 | register_as_ext3(); |
2035e776 | 4979 | register_as_ext2(); |
03010a33 | 4980 | err = register_filesystem(&ext4_fs_type); |
ac27a0ec DK |
4981 | if (err) |
4982 | goto out; | |
bfff6873 LC |
4983 | |
4984 | ext4_li_info = NULL; | |
4985 | mutex_init(&ext4_li_mtx); | |
ac27a0ec DK |
4986 | return 0; |
4987 | out: | |
24b58424 TT |
4988 | unregister_as_ext2(); |
4989 | unregister_as_ext3(); | |
ac27a0ec DK |
4990 | destroy_inodecache(); |
4991 | out1: | |
5dabfc78 | 4992 | ext4_exit_xattr(); |
c9de560d | 4993 | out2: |
5dabfc78 | 4994 | ext4_exit_mballoc(); |
6fd058f7 | 4995 | out3: |
8f021222 | 4996 | ext4_exit_feat_adverts(); |
dd68314c | 4997 | out4: |
6fd058f7 | 4998 | remove_proc_entry("fs/ext4", NULL); |
dd68314c | 4999 | out5: |
6fd058f7 | 5000 | kset_unregister(ext4_kset); |
dd68314c | 5001 | out6: |
5dabfc78 | 5002 | ext4_exit_system_zone(); |
dd68314c | 5003 | out7: |
5dabfc78 | 5004 | ext4_exit_pageio(); |
ac27a0ec DK |
5005 | return err; |
5006 | } | |
5007 | ||
5dabfc78 | 5008 | static void __exit ext4_exit_fs(void) |
ac27a0ec | 5009 | { |
bfff6873 | 5010 | ext4_destroy_lazyinit_thread(); |
24b58424 TT |
5011 | unregister_as_ext2(); |
5012 | unregister_as_ext3(); | |
03010a33 | 5013 | unregister_filesystem(&ext4_fs_type); |
ac27a0ec | 5014 | destroy_inodecache(); |
5dabfc78 TT |
5015 | ext4_exit_xattr(); |
5016 | ext4_exit_mballoc(); | |
8f021222 | 5017 | ext4_exit_feat_adverts(); |
9f6200bb | 5018 | remove_proc_entry("fs/ext4", NULL); |
3197ebdb | 5019 | kset_unregister(ext4_kset); |
5dabfc78 TT |
5020 | ext4_exit_system_zone(); |
5021 | ext4_exit_pageio(); | |
ac27a0ec DK |
5022 | } |
5023 | ||
5024 | MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); | |
83982b6f | 5025 | MODULE_DESCRIPTION("Fourth Extended Filesystem"); |
ac27a0ec | 5026 | MODULE_LICENSE("GPL"); |
5dabfc78 TT |
5027 | module_init(ext4_init_fs) |
5028 | module_exit(ext4_exit_fs) |