]>
Commit | Line | Data |
---|---|---|
ac27a0ec | 1 | /* |
617ba13b | 2 | * linux/fs/ext4/super.c |
ac27a0ec DK |
3 | * |
4 | * Copyright (C) 1992, 1993, 1994, 1995 | |
5 | * Remy Card (card@masi.ibp.fr) | |
6 | * Laboratoire MASI - Institut Blaise Pascal | |
7 | * Universite Pierre et Marie Curie (Paris VI) | |
8 | * | |
9 | * from | |
10 | * | |
11 | * linux/fs/minix/inode.c | |
12 | * | |
13 | * Copyright (C) 1991, 1992 Linus Torvalds | |
14 | * | |
15 | * Big-endian to little-endian byte-swapping/bitmaps by | |
16 | * David S. Miller (davem@caip.rutgers.edu), 1995 | |
17 | */ | |
18 | ||
19 | #include <linux/module.h> | |
20 | #include <linux/string.h> | |
21 | #include <linux/fs.h> | |
22 | #include <linux/time.h> | |
c5ca7c76 | 23 | #include <linux/vmalloc.h> |
dab291af | 24 | #include <linux/jbd2.h> |
ac27a0ec DK |
25 | #include <linux/slab.h> |
26 | #include <linux/init.h> | |
27 | #include <linux/blkdev.h> | |
28 | #include <linux/parser.h> | |
ac27a0ec | 29 | #include <linux/buffer_head.h> |
a5694255 | 30 | #include <linux/exportfs.h> |
ac27a0ec DK |
31 | #include <linux/vfs.h> |
32 | #include <linux/random.h> | |
33 | #include <linux/mount.h> | |
34 | #include <linux/namei.h> | |
35 | #include <linux/quotaops.h> | |
36 | #include <linux/seq_file.h> | |
9f6200bb | 37 | #include <linux/proc_fs.h> |
3197ebdb | 38 | #include <linux/ctype.h> |
1330593e | 39 | #include <linux/log2.h> |
717d50e4 | 40 | #include <linux/crc16.h> |
ac27a0ec DK |
41 | #include <asm/uaccess.h> |
42 | ||
bfff6873 LC |
43 | #include <linux/kthread.h> |
44 | #include <linux/freezer.h> | |
45 | ||
3dcf5451 CH |
46 | #include "ext4.h" |
47 | #include "ext4_jbd2.h" | |
ac27a0ec DK |
48 | #include "xattr.h" |
49 | #include "acl.h" | |
3661d286 | 50 | #include "mballoc.h" |
ac27a0ec | 51 | |
9bffad1e TT |
52 | #define CREATE_TRACE_POINTS |
53 | #include <trace/events/ext4.h> | |
54 | ||
1f109d5a | 55 | static struct proc_dir_entry *ext4_proc_root; |
3197ebdb | 56 | static struct kset *ext4_kset; |
bfff6873 LC |
57 | struct ext4_lazy_init *ext4_li_info; |
58 | struct mutex ext4_li_mtx; | |
857ac889 | 59 | struct ext4_features *ext4_feat; |
9f6200bb | 60 | |
617ba13b | 61 | static int ext4_load_journal(struct super_block *, struct ext4_super_block *, |
ac27a0ec | 62 | unsigned long journal_devnum); |
e2d67052 | 63 | static int ext4_commit_super(struct super_block *sb, int sync); |
2b2d6d01 TT |
64 | static void ext4_mark_recovery_complete(struct super_block *sb, |
65 | struct ext4_super_block *es); | |
66 | static void ext4_clear_journal_err(struct super_block *sb, | |
67 | struct ext4_super_block *es); | |
617ba13b | 68 | static int ext4_sync_fs(struct super_block *sb, int wait); |
2b2d6d01 | 69 | static const char *ext4_decode_error(struct super_block *sb, int errno, |
ac27a0ec | 70 | char nbuf[16]); |
2b2d6d01 TT |
71 | static int ext4_remount(struct super_block *sb, int *flags, char *data); |
72 | static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); | |
c4be0c1d | 73 | static int ext4_unfreeze(struct super_block *sb); |
2b2d6d01 | 74 | static void ext4_write_super(struct super_block *sb); |
c4be0c1d | 75 | static int ext4_freeze(struct super_block *sb); |
152a0836 AV |
76 | static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, |
77 | const char *dev_name, void *data); | |
bfff6873 LC |
78 | static void ext4_destroy_lazyinit_thread(void); |
79 | static void ext4_unregister_li_request(struct super_block *sb); | |
ac27a0ec | 80 | |
ba69f9ab JK |
81 | #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) |
82 | static struct file_system_type ext3_fs_type = { | |
83 | .owner = THIS_MODULE, | |
84 | .name = "ext3", | |
152a0836 | 85 | .mount = ext4_mount, |
ba69f9ab JK |
86 | .kill_sb = kill_block_super, |
87 | .fs_flags = FS_REQUIRES_DEV, | |
88 | }; | |
89 | #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) | |
90 | #else | |
91 | #define IS_EXT3_SB(sb) (0) | |
92 | #endif | |
bd81d8ee | 93 | |
8fadc143 AR |
94 | ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, |
95 | struct ext4_group_desc *bg) | |
bd81d8ee | 96 | { |
3a14589c | 97 | return le32_to_cpu(bg->bg_block_bitmap_lo) | |
8fadc143 | 98 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 99 | (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); |
bd81d8ee LV |
100 | } |
101 | ||
8fadc143 AR |
102 | ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, |
103 | struct ext4_group_desc *bg) | |
bd81d8ee | 104 | { |
5272f837 | 105 | return le32_to_cpu(bg->bg_inode_bitmap_lo) | |
8fadc143 | 106 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 107 | (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); |
bd81d8ee LV |
108 | } |
109 | ||
8fadc143 AR |
110 | ext4_fsblk_t ext4_inode_table(struct super_block *sb, |
111 | struct ext4_group_desc *bg) | |
bd81d8ee | 112 | { |
5272f837 | 113 | return le32_to_cpu(bg->bg_inode_table_lo) | |
8fadc143 | 114 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 115 | (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); |
bd81d8ee LV |
116 | } |
117 | ||
560671a0 AK |
118 | __u32 ext4_free_blks_count(struct super_block *sb, |
119 | struct ext4_group_desc *bg) | |
120 | { | |
121 | return le16_to_cpu(bg->bg_free_blocks_count_lo) | | |
122 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 123 | (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); |
560671a0 AK |
124 | } |
125 | ||
126 | __u32 ext4_free_inodes_count(struct super_block *sb, | |
127 | struct ext4_group_desc *bg) | |
128 | { | |
129 | return le16_to_cpu(bg->bg_free_inodes_count_lo) | | |
130 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 131 | (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); |
560671a0 AK |
132 | } |
133 | ||
134 | __u32 ext4_used_dirs_count(struct super_block *sb, | |
135 | struct ext4_group_desc *bg) | |
136 | { | |
137 | return le16_to_cpu(bg->bg_used_dirs_count_lo) | | |
138 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 139 | (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); |
560671a0 AK |
140 | } |
141 | ||
142 | __u32 ext4_itable_unused_count(struct super_block *sb, | |
143 | struct ext4_group_desc *bg) | |
144 | { | |
145 | return le16_to_cpu(bg->bg_itable_unused_lo) | | |
146 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 147 | (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); |
560671a0 AK |
148 | } |
149 | ||
8fadc143 AR |
150 | void ext4_block_bitmap_set(struct super_block *sb, |
151 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 152 | { |
3a14589c | 153 | bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
154 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
155 | bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
156 | } |
157 | ||
8fadc143 AR |
158 | void ext4_inode_bitmap_set(struct super_block *sb, |
159 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 160 | { |
5272f837 | 161 | bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
162 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
163 | bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
164 | } |
165 | ||
8fadc143 AR |
166 | void ext4_inode_table_set(struct super_block *sb, |
167 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 168 | { |
5272f837 | 169 | bg->bg_inode_table_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
170 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
171 | bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
172 | } |
173 | ||
560671a0 AK |
174 | void ext4_free_blks_set(struct super_block *sb, |
175 | struct ext4_group_desc *bg, __u32 count) | |
176 | { | |
177 | bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); | |
178 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
179 | bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); | |
180 | } | |
181 | ||
182 | void ext4_free_inodes_set(struct super_block *sb, | |
183 | struct ext4_group_desc *bg, __u32 count) | |
184 | { | |
185 | bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); | |
186 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
187 | bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); | |
188 | } | |
189 | ||
190 | void ext4_used_dirs_set(struct super_block *sb, | |
191 | struct ext4_group_desc *bg, __u32 count) | |
192 | { | |
193 | bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); | |
194 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
195 | bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); | |
196 | } | |
197 | ||
198 | void ext4_itable_unused_set(struct super_block *sb, | |
199 | struct ext4_group_desc *bg, __u32 count) | |
200 | { | |
201 | bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); | |
202 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
203 | bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); | |
204 | } | |
205 | ||
d3d1faf6 CW |
206 | |
207 | /* Just increment the non-pointer handle value */ | |
208 | static handle_t *ext4_get_nojournal(void) | |
209 | { | |
210 | handle_t *handle = current->journal_info; | |
211 | unsigned long ref_cnt = (unsigned long)handle; | |
212 | ||
213 | BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT); | |
214 | ||
215 | ref_cnt++; | |
216 | handle = (handle_t *)ref_cnt; | |
217 | ||
218 | current->journal_info = handle; | |
219 | return handle; | |
220 | } | |
221 | ||
222 | ||
223 | /* Decrement the non-pointer handle value */ | |
224 | static void ext4_put_nojournal(handle_t *handle) | |
225 | { | |
226 | unsigned long ref_cnt = (unsigned long)handle; | |
227 | ||
228 | BUG_ON(ref_cnt == 0); | |
229 | ||
230 | ref_cnt--; | |
231 | handle = (handle_t *)ref_cnt; | |
232 | ||
233 | current->journal_info = handle; | |
234 | } | |
235 | ||
ac27a0ec | 236 | /* |
dab291af | 237 | * Wrappers for jbd2_journal_start/end. |
ac27a0ec DK |
238 | * |
239 | * The only special thing we need to do here is to make sure that all | |
240 | * journal_end calls result in the superblock being marked dirty, so | |
241 | * that sync() will call the filesystem's write_super callback if | |
242 | * appropriate. | |
243 | */ | |
617ba13b | 244 | handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks) |
ac27a0ec DK |
245 | { |
246 | journal_t *journal; | |
247 | ||
248 | if (sb->s_flags & MS_RDONLY) | |
249 | return ERR_PTR(-EROFS); | |
250 | ||
437f88cc | 251 | vfs_check_frozen(sb, SB_FREEZE_TRANS); |
ac27a0ec DK |
252 | /* Special case here: if the journal has aborted behind our |
253 | * backs (eg. EIO in the commit thread), then we still need to | |
254 | * take the FS itself readonly cleanly. */ | |
617ba13b | 255 | journal = EXT4_SB(sb)->s_journal; |
0390131b FM |
256 | if (journal) { |
257 | if (is_journal_aborted(journal)) { | |
c67d859e | 258 | ext4_abort(sb, "Detected aborted journal"); |
0390131b FM |
259 | return ERR_PTR(-EROFS); |
260 | } | |
261 | return jbd2_journal_start(journal, nblocks); | |
ac27a0ec | 262 | } |
d3d1faf6 | 263 | return ext4_get_nojournal(); |
ac27a0ec DK |
264 | } |
265 | ||
266 | /* | |
267 | * The only special thing we need to do here is to make sure that all | |
dab291af | 268 | * jbd2_journal_stop calls result in the superblock being marked dirty, so |
ac27a0ec DK |
269 | * that sync() will call the filesystem's write_super callback if |
270 | * appropriate. | |
271 | */ | |
c398eda0 | 272 | int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle) |
ac27a0ec DK |
273 | { |
274 | struct super_block *sb; | |
275 | int err; | |
276 | int rc; | |
277 | ||
0390131b | 278 | if (!ext4_handle_valid(handle)) { |
d3d1faf6 | 279 | ext4_put_nojournal(handle); |
0390131b FM |
280 | return 0; |
281 | } | |
ac27a0ec DK |
282 | sb = handle->h_transaction->t_journal->j_private; |
283 | err = handle->h_err; | |
dab291af | 284 | rc = jbd2_journal_stop(handle); |
ac27a0ec DK |
285 | |
286 | if (!err) | |
287 | err = rc; | |
288 | if (err) | |
c398eda0 | 289 | __ext4_std_error(sb, where, line, err); |
ac27a0ec DK |
290 | return err; |
291 | } | |
292 | ||
90c7201b TT |
293 | void ext4_journal_abort_handle(const char *caller, unsigned int line, |
294 | const char *err_fn, struct buffer_head *bh, | |
295 | handle_t *handle, int err) | |
ac27a0ec DK |
296 | { |
297 | char nbuf[16]; | |
617ba13b | 298 | const char *errstr = ext4_decode_error(NULL, err, nbuf); |
ac27a0ec | 299 | |
0390131b FM |
300 | BUG_ON(!ext4_handle_valid(handle)); |
301 | ||
ac27a0ec DK |
302 | if (bh) |
303 | BUFFER_TRACE(bh, "abort"); | |
304 | ||
305 | if (!handle->h_err) | |
306 | handle->h_err = err; | |
307 | ||
308 | if (is_handle_aborted(handle)) | |
309 | return; | |
310 | ||
90c7201b TT |
311 | printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n", |
312 | caller, line, errstr, err_fn); | |
ac27a0ec | 313 | |
dab291af | 314 | jbd2_journal_abort_handle(handle); |
ac27a0ec DK |
315 | } |
316 | ||
1c13d5c0 TT |
317 | static void __save_error_info(struct super_block *sb, const char *func, |
318 | unsigned int line) | |
319 | { | |
320 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; | |
321 | ||
322 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; | |
323 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
324 | es->s_last_error_time = cpu_to_le32(get_seconds()); | |
325 | strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func)); | |
326 | es->s_last_error_line = cpu_to_le32(line); | |
327 | if (!es->s_first_error_time) { | |
328 | es->s_first_error_time = es->s_last_error_time; | |
329 | strncpy(es->s_first_error_func, func, | |
330 | sizeof(es->s_first_error_func)); | |
331 | es->s_first_error_line = cpu_to_le32(line); | |
332 | es->s_first_error_ino = es->s_last_error_ino; | |
333 | es->s_first_error_block = es->s_last_error_block; | |
334 | } | |
66e61a9e TT |
335 | /* |
336 | * Start the daily error reporting function if it hasn't been | |
337 | * started already | |
338 | */ | |
339 | if (!es->s_error_count) | |
340 | mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ); | |
1c13d5c0 TT |
341 | es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1); |
342 | } | |
343 | ||
344 | static void save_error_info(struct super_block *sb, const char *func, | |
345 | unsigned int line) | |
346 | { | |
347 | __save_error_info(sb, func, line); | |
348 | ext4_commit_super(sb, 1); | |
349 | } | |
350 | ||
351 | ||
ac27a0ec DK |
352 | /* Deal with the reporting of failure conditions on a filesystem such as |
353 | * inconsistencies detected or read IO failures. | |
354 | * | |
355 | * On ext2, we can store the error state of the filesystem in the | |
617ba13b | 356 | * superblock. That is not possible on ext4, because we may have other |
ac27a0ec DK |
357 | * write ordering constraints on the superblock which prevent us from |
358 | * writing it out straight away; and given that the journal is about to | |
359 | * be aborted, we can't rely on the current, or future, transactions to | |
360 | * write out the superblock safely. | |
361 | * | |
dab291af | 362 | * We'll just use the jbd2_journal_abort() error code to record an error in |
d6b198bc | 363 | * the journal instead. On recovery, the journal will complain about |
ac27a0ec DK |
364 | * that error until we've noted it down and cleared it. |
365 | */ | |
366 | ||
617ba13b | 367 | static void ext4_handle_error(struct super_block *sb) |
ac27a0ec | 368 | { |
ac27a0ec DK |
369 | if (sb->s_flags & MS_RDONLY) |
370 | return; | |
371 | ||
2b2d6d01 | 372 | if (!test_opt(sb, ERRORS_CONT)) { |
617ba13b | 373 | journal_t *journal = EXT4_SB(sb)->s_journal; |
ac27a0ec | 374 | |
4ab2f15b | 375 | EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; |
ac27a0ec | 376 | if (journal) |
dab291af | 377 | jbd2_journal_abort(journal, -EIO); |
ac27a0ec | 378 | } |
2b2d6d01 | 379 | if (test_opt(sb, ERRORS_RO)) { |
b31e1552 | 380 | ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); |
ac27a0ec DK |
381 | sb->s_flags |= MS_RDONLY; |
382 | } | |
ac27a0ec | 383 | if (test_opt(sb, ERRORS_PANIC)) |
617ba13b | 384 | panic("EXT4-fs (device %s): panic forced after error\n", |
ac27a0ec DK |
385 | sb->s_id); |
386 | } | |
387 | ||
12062ddd | 388 | void __ext4_error(struct super_block *sb, const char *function, |
c398eda0 | 389 | unsigned int line, const char *fmt, ...) |
ac27a0ec | 390 | { |
0ff2ea7d | 391 | struct va_format vaf; |
ac27a0ec DK |
392 | va_list args; |
393 | ||
394 | va_start(args, fmt); | |
0ff2ea7d JP |
395 | vaf.fmt = fmt; |
396 | vaf.va = &args; | |
397 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", | |
398 | sb->s_id, function, line, current->comm, &vaf); | |
ac27a0ec DK |
399 | va_end(args); |
400 | ||
617ba13b | 401 | ext4_handle_error(sb); |
ac27a0ec DK |
402 | } |
403 | ||
c398eda0 TT |
404 | void ext4_error_inode(struct inode *inode, const char *function, |
405 | unsigned int line, ext4_fsblk_t block, | |
273df556 FM |
406 | const char *fmt, ...) |
407 | { | |
408 | va_list args; | |
1c13d5c0 | 409 | struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es; |
273df556 | 410 | |
1c13d5c0 TT |
411 | es->s_last_error_ino = cpu_to_le32(inode->i_ino); |
412 | es->s_last_error_block = cpu_to_le64(block); | |
413 | save_error_info(inode->i_sb, function, line); | |
273df556 | 414 | va_start(args, fmt); |
c398eda0 TT |
415 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ", |
416 | inode->i_sb->s_id, function, line, inode->i_ino); | |
417 | if (block) | |
418 | printk("block %llu: ", block); | |
419 | printk("comm %s: ", current->comm); | |
273df556 FM |
420 | vprintk(fmt, args); |
421 | printk("\n"); | |
422 | va_end(args); | |
423 | ||
424 | ext4_handle_error(inode->i_sb); | |
425 | } | |
426 | ||
c398eda0 TT |
427 | void ext4_error_file(struct file *file, const char *function, |
428 | unsigned int line, const char *fmt, ...) | |
273df556 FM |
429 | { |
430 | va_list args; | |
1c13d5c0 | 431 | struct ext4_super_block *es; |
273df556 FM |
432 | struct inode *inode = file->f_dentry->d_inode; |
433 | char pathname[80], *path; | |
434 | ||
1c13d5c0 TT |
435 | es = EXT4_SB(inode->i_sb)->s_es; |
436 | es->s_last_error_ino = cpu_to_le32(inode->i_ino); | |
437 | save_error_info(inode->i_sb, function, line); | |
273df556 FM |
438 | va_start(args, fmt); |
439 | path = d_path(&(file->f_path), pathname, sizeof(pathname)); | |
440 | if (!path) | |
441 | path = "(unknown)"; | |
442 | printk(KERN_CRIT | |
c398eda0 TT |
443 | "EXT4-fs error (device %s): %s:%d: inode #%lu " |
444 | "(comm %s path %s): ", | |
445 | inode->i_sb->s_id, function, line, inode->i_ino, | |
446 | current->comm, path); | |
273df556 FM |
447 | vprintk(fmt, args); |
448 | printk("\n"); | |
449 | va_end(args); | |
450 | ||
451 | ext4_handle_error(inode->i_sb); | |
452 | } | |
453 | ||
2b2d6d01 | 454 | static const char *ext4_decode_error(struct super_block *sb, int errno, |
ac27a0ec DK |
455 | char nbuf[16]) |
456 | { | |
457 | char *errstr = NULL; | |
458 | ||
459 | switch (errno) { | |
460 | case -EIO: | |
461 | errstr = "IO failure"; | |
462 | break; | |
463 | case -ENOMEM: | |
464 | errstr = "Out of memory"; | |
465 | break; | |
466 | case -EROFS: | |
78f1ddbb TT |
467 | if (!sb || (EXT4_SB(sb)->s_journal && |
468 | EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) | |
ac27a0ec DK |
469 | errstr = "Journal has aborted"; |
470 | else | |
471 | errstr = "Readonly filesystem"; | |
472 | break; | |
473 | default: | |
474 | /* If the caller passed in an extra buffer for unknown | |
475 | * errors, textualise them now. Else we just return | |
476 | * NULL. */ | |
477 | if (nbuf) { | |
478 | /* Check for truncated error codes... */ | |
479 | if (snprintf(nbuf, 16, "error %d", -errno) >= 0) | |
480 | errstr = nbuf; | |
481 | } | |
482 | break; | |
483 | } | |
484 | ||
485 | return errstr; | |
486 | } | |
487 | ||
617ba13b | 488 | /* __ext4_std_error decodes expected errors from journaling functions |
ac27a0ec DK |
489 | * automatically and invokes the appropriate error response. */ |
490 | ||
c398eda0 TT |
491 | void __ext4_std_error(struct super_block *sb, const char *function, |
492 | unsigned int line, int errno) | |
ac27a0ec DK |
493 | { |
494 | char nbuf[16]; | |
495 | const char *errstr; | |
496 | ||
497 | /* Special case: if the error is EROFS, and we're not already | |
498 | * inside a transaction, then there's really no point in logging | |
499 | * an error. */ | |
500 | if (errno == -EROFS && journal_current_handle() == NULL && | |
501 | (sb->s_flags & MS_RDONLY)) | |
502 | return; | |
503 | ||
617ba13b | 504 | errstr = ext4_decode_error(sb, errno, nbuf); |
c398eda0 TT |
505 | printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", |
506 | sb->s_id, function, line, errstr); | |
1c13d5c0 | 507 | save_error_info(sb, function, line); |
ac27a0ec | 508 | |
617ba13b | 509 | ext4_handle_error(sb); |
ac27a0ec DK |
510 | } |
511 | ||
512 | /* | |
617ba13b | 513 | * ext4_abort is a much stronger failure handler than ext4_error. The |
ac27a0ec DK |
514 | * abort function may be used to deal with unrecoverable failures such |
515 | * as journal IO errors or ENOMEM at a critical moment in log management. | |
516 | * | |
517 | * We unconditionally force the filesystem into an ABORT|READONLY state, | |
518 | * unless the error response on the fs has been set to panic in which | |
519 | * case we take the easy way out and panic immediately. | |
520 | */ | |
521 | ||
c67d859e | 522 | void __ext4_abort(struct super_block *sb, const char *function, |
c398eda0 | 523 | unsigned int line, const char *fmt, ...) |
ac27a0ec DK |
524 | { |
525 | va_list args; | |
526 | ||
1c13d5c0 | 527 | save_error_info(sb, function, line); |
ac27a0ec | 528 | va_start(args, fmt); |
c398eda0 TT |
529 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id, |
530 | function, line); | |
ac27a0ec DK |
531 | vprintk(fmt, args); |
532 | printk("\n"); | |
533 | va_end(args); | |
534 | ||
1c13d5c0 TT |
535 | if ((sb->s_flags & MS_RDONLY) == 0) { |
536 | ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); | |
537 | sb->s_flags |= MS_RDONLY; | |
538 | EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; | |
539 | if (EXT4_SB(sb)->s_journal) | |
540 | jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); | |
541 | save_error_info(sb, function, line); | |
542 | } | |
ac27a0ec | 543 | if (test_opt(sb, ERRORS_PANIC)) |
617ba13b | 544 | panic("EXT4-fs panic from previous error\n"); |
ac27a0ec DK |
545 | } |
546 | ||
0ff2ea7d | 547 | void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...) |
b31e1552 | 548 | { |
0ff2ea7d | 549 | struct va_format vaf; |
b31e1552 ES |
550 | va_list args; |
551 | ||
552 | va_start(args, fmt); | |
0ff2ea7d JP |
553 | vaf.fmt = fmt; |
554 | vaf.va = &args; | |
555 | printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); | |
b31e1552 ES |
556 | va_end(args); |
557 | } | |
558 | ||
12062ddd | 559 | void __ext4_warning(struct super_block *sb, const char *function, |
c398eda0 | 560 | unsigned int line, const char *fmt, ...) |
ac27a0ec | 561 | { |
0ff2ea7d | 562 | struct va_format vaf; |
ac27a0ec DK |
563 | va_list args; |
564 | ||
565 | va_start(args, fmt); | |
0ff2ea7d JP |
566 | vaf.fmt = fmt; |
567 | vaf.va = &args; | |
568 | printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", | |
569 | sb->s_id, function, line, &vaf); | |
ac27a0ec DK |
570 | va_end(args); |
571 | } | |
572 | ||
e29136f8 TT |
573 | void __ext4_grp_locked_error(const char *function, unsigned int line, |
574 | struct super_block *sb, ext4_group_t grp, | |
575 | unsigned long ino, ext4_fsblk_t block, | |
576 | const char *fmt, ...) | |
5d1b1b3f AK |
577 | __releases(bitlock) |
578 | __acquires(bitlock) | |
579 | { | |
0ff2ea7d | 580 | struct va_format vaf; |
5d1b1b3f AK |
581 | va_list args; |
582 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; | |
583 | ||
1c13d5c0 TT |
584 | es->s_last_error_ino = cpu_to_le32(ino); |
585 | es->s_last_error_block = cpu_to_le64(block); | |
586 | __save_error_info(sb, function, line); | |
0ff2ea7d | 587 | |
5d1b1b3f | 588 | va_start(args, fmt); |
0ff2ea7d JP |
589 | |
590 | vaf.fmt = fmt; | |
591 | vaf.va = &args; | |
e29136f8 TT |
592 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u", |
593 | sb->s_id, function, line, grp); | |
594 | if (ino) | |
0ff2ea7d | 595 | printk(KERN_CONT "inode %lu: ", ino); |
e29136f8 | 596 | if (block) |
0ff2ea7d JP |
597 | printk(KERN_CONT "block %llu:", (unsigned long long) block); |
598 | printk(KERN_CONT "%pV\n", &vaf); | |
5d1b1b3f AK |
599 | va_end(args); |
600 | ||
601 | if (test_opt(sb, ERRORS_CONT)) { | |
e2d67052 | 602 | ext4_commit_super(sb, 0); |
5d1b1b3f AK |
603 | return; |
604 | } | |
1c13d5c0 | 605 | |
5d1b1b3f AK |
606 | ext4_unlock_group(sb, grp); |
607 | ext4_handle_error(sb); | |
608 | /* | |
609 | * We only get here in the ERRORS_RO case; relocking the group | |
610 | * may be dangerous, but nothing bad will happen since the | |
611 | * filesystem will have already been marked read/only and the | |
612 | * journal has been aborted. We return 1 as a hint to callers | |
613 | * who might what to use the return value from | |
614 | * ext4_grp_locked_error() to distinguish beween the | |
615 | * ERRORS_CONT and ERRORS_RO case, and perhaps return more | |
616 | * aggressively from the ext4 function in question, with a | |
617 | * more appropriate error code. | |
618 | */ | |
619 | ext4_lock_group(sb, grp); | |
620 | return; | |
621 | } | |
622 | ||
617ba13b | 623 | void ext4_update_dynamic_rev(struct super_block *sb) |
ac27a0ec | 624 | { |
617ba13b | 625 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
ac27a0ec | 626 | |
617ba13b | 627 | if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) |
ac27a0ec DK |
628 | return; |
629 | ||
12062ddd | 630 | ext4_warning(sb, |
ac27a0ec DK |
631 | "updating to rev %d because of new feature flag, " |
632 | "running e2fsck is recommended", | |
617ba13b | 633 | EXT4_DYNAMIC_REV); |
ac27a0ec | 634 | |
617ba13b MC |
635 | es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); |
636 | es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); | |
637 | es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); | |
ac27a0ec DK |
638 | /* leave es->s_feature_*compat flags alone */ |
639 | /* es->s_uuid will be set by e2fsck if empty */ | |
640 | ||
641 | /* | |
642 | * The rest of the superblock fields should be zero, and if not it | |
643 | * means they are likely already in use, so leave them alone. We | |
644 | * can leave it up to e2fsck to clean up any inconsistencies there. | |
645 | */ | |
646 | } | |
647 | ||
648 | /* | |
649 | * Open the external journal device | |
650 | */ | |
b31e1552 | 651 | static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) |
ac27a0ec DK |
652 | { |
653 | struct block_device *bdev; | |
654 | char b[BDEVNAME_SIZE]; | |
655 | ||
656 | bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE); | |
657 | if (IS_ERR(bdev)) | |
658 | goto fail; | |
659 | return bdev; | |
660 | ||
661 | fail: | |
b31e1552 | 662 | ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld", |
ac27a0ec DK |
663 | __bdevname(dev, b), PTR_ERR(bdev)); |
664 | return NULL; | |
665 | } | |
666 | ||
667 | /* | |
668 | * Release the journal device | |
669 | */ | |
617ba13b | 670 | static int ext4_blkdev_put(struct block_device *bdev) |
ac27a0ec DK |
671 | { |
672 | bd_release(bdev); | |
9a1c3542 | 673 | return blkdev_put(bdev, FMODE_READ|FMODE_WRITE); |
ac27a0ec DK |
674 | } |
675 | ||
617ba13b | 676 | static int ext4_blkdev_remove(struct ext4_sb_info *sbi) |
ac27a0ec DK |
677 | { |
678 | struct block_device *bdev; | |
679 | int ret = -ENODEV; | |
680 | ||
681 | bdev = sbi->journal_bdev; | |
682 | if (bdev) { | |
617ba13b | 683 | ret = ext4_blkdev_put(bdev); |
ac27a0ec DK |
684 | sbi->journal_bdev = NULL; |
685 | } | |
686 | return ret; | |
687 | } | |
688 | ||
689 | static inline struct inode *orphan_list_entry(struct list_head *l) | |
690 | { | |
617ba13b | 691 | return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; |
ac27a0ec DK |
692 | } |
693 | ||
617ba13b | 694 | static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) |
ac27a0ec DK |
695 | { |
696 | struct list_head *l; | |
697 | ||
b31e1552 ES |
698 | ext4_msg(sb, KERN_ERR, "sb orphan head is %d", |
699 | le32_to_cpu(sbi->s_es->s_last_orphan)); | |
ac27a0ec DK |
700 | |
701 | printk(KERN_ERR "sb_info orphan list:\n"); | |
702 | list_for_each(l, &sbi->s_orphan) { | |
703 | struct inode *inode = orphan_list_entry(l); | |
704 | printk(KERN_ERR " " | |
705 | "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", | |
706 | inode->i_sb->s_id, inode->i_ino, inode, | |
707 | inode->i_mode, inode->i_nlink, | |
708 | NEXT_ORPHAN(inode)); | |
709 | } | |
710 | } | |
711 | ||
2b2d6d01 | 712 | static void ext4_put_super(struct super_block *sb) |
ac27a0ec | 713 | { |
617ba13b MC |
714 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
715 | struct ext4_super_block *es = sbi->s_es; | |
ef2cabf7 | 716 | int i, err; |
ac27a0ec | 717 | |
857ac889 | 718 | ext4_unregister_li_request(sb); |
e0ccfd95 CH |
719 | dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED); |
720 | ||
4c0425ff MC |
721 | flush_workqueue(sbi->dio_unwritten_wq); |
722 | destroy_workqueue(sbi->dio_unwritten_wq); | |
723 | ||
a9e220f8 | 724 | lock_super(sb); |
8c85e125 | 725 | if (sb->s_dirt) |
ebc1ac16 | 726 | ext4_commit_super(sb, 1); |
8c85e125 | 727 | |
0390131b FM |
728 | if (sbi->s_journal) { |
729 | err = jbd2_journal_destroy(sbi->s_journal); | |
730 | sbi->s_journal = NULL; | |
731 | if (err < 0) | |
c67d859e | 732 | ext4_abort(sb, "Couldn't clean up the journal"); |
0390131b | 733 | } |
d4edac31 | 734 | |
a1c6c569 | 735 | del_timer(&sbi->s_err_report); |
d4edac31 JB |
736 | ext4_release_system_zone(sb); |
737 | ext4_mb_release(sb); | |
738 | ext4_ext_release(sb); | |
739 | ext4_xattr_put_super(sb); | |
740 | ||
ac27a0ec | 741 | if (!(sb->s_flags & MS_RDONLY)) { |
617ba13b | 742 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); |
ac27a0ec | 743 | es->s_state = cpu_to_le16(sbi->s_mount_state); |
e2d67052 | 744 | ext4_commit_super(sb, 1); |
ac27a0ec | 745 | } |
240799cd | 746 | if (sbi->s_proc) { |
9f6200bb | 747 | remove_proc_entry(sb->s_id, ext4_proc_root); |
240799cd | 748 | } |
3197ebdb | 749 | kobject_del(&sbi->s_kobj); |
ac27a0ec DK |
750 | |
751 | for (i = 0; i < sbi->s_gdb_count; i++) | |
752 | brelse(sbi->s_group_desc[i]); | |
753 | kfree(sbi->s_group_desc); | |
c5ca7c76 TT |
754 | if (is_vmalloc_addr(sbi->s_flex_groups)) |
755 | vfree(sbi->s_flex_groups); | |
756 | else | |
757 | kfree(sbi->s_flex_groups); | |
ac27a0ec DK |
758 | percpu_counter_destroy(&sbi->s_freeblocks_counter); |
759 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
760 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
6bc6e63f | 761 | percpu_counter_destroy(&sbi->s_dirtyblocks_counter); |
ac27a0ec DK |
762 | brelse(sbi->s_sbh); |
763 | #ifdef CONFIG_QUOTA | |
764 | for (i = 0; i < MAXQUOTAS; i++) | |
765 | kfree(sbi->s_qf_names[i]); | |
766 | #endif | |
767 | ||
768 | /* Debugging code just in case the in-memory inode orphan list | |
769 | * isn't empty. The on-disk one can be non-empty if we've | |
770 | * detected an error and taken the fs readonly, but the | |
771 | * in-memory list had better be clean by this point. */ | |
772 | if (!list_empty(&sbi->s_orphan)) | |
773 | dump_orphan_list(sb, sbi); | |
774 | J_ASSERT(list_empty(&sbi->s_orphan)); | |
775 | ||
f98393a6 | 776 | invalidate_bdev(sb->s_bdev); |
ac27a0ec DK |
777 | if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { |
778 | /* | |
779 | * Invalidate the journal device's buffers. We don't want them | |
780 | * floating about in memory - the physical journal device may | |
781 | * hotswapped, and it breaks the `ro-after' testing code. | |
782 | */ | |
783 | sync_blockdev(sbi->journal_bdev); | |
f98393a6 | 784 | invalidate_bdev(sbi->journal_bdev); |
617ba13b | 785 | ext4_blkdev_remove(sbi); |
ac27a0ec DK |
786 | } |
787 | sb->s_fs_info = NULL; | |
3197ebdb TT |
788 | /* |
789 | * Now that we are completely done shutting down the | |
790 | * superblock, we need to actually destroy the kobject. | |
791 | */ | |
3197ebdb TT |
792 | unlock_super(sb); |
793 | kobject_put(&sbi->s_kobj); | |
794 | wait_for_completion(&sbi->s_kobj_unregister); | |
705895b6 | 795 | kfree(sbi->s_blockgroup_lock); |
ac27a0ec | 796 | kfree(sbi); |
ac27a0ec DK |
797 | } |
798 | ||
e18b890b | 799 | static struct kmem_cache *ext4_inode_cachep; |
ac27a0ec DK |
800 | |
801 | /* | |
802 | * Called inside transaction, so use GFP_NOFS | |
803 | */ | |
617ba13b | 804 | static struct inode *ext4_alloc_inode(struct super_block *sb) |
ac27a0ec | 805 | { |
617ba13b | 806 | struct ext4_inode_info *ei; |
ac27a0ec | 807 | |
e6b4f8da | 808 | ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); |
ac27a0ec DK |
809 | if (!ei) |
810 | return NULL; | |
0b8e58a1 | 811 | |
ac27a0ec | 812 | ei->vfs_inode.i_version = 1; |
91246c00 | 813 | ei->vfs_inode.i_data.writeback_index = 0; |
a86c6181 | 814 | memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache)); |
c9de560d AT |
815 | INIT_LIST_HEAD(&ei->i_prealloc_list); |
816 | spin_lock_init(&ei->i_prealloc_lock); | |
0390131b FM |
817 | /* |
818 | * Note: We can be called before EXT4_SB(sb)->s_journal is set, | |
819 | * therefore it can be null here. Don't check it, just initialize | |
820 | * jinode. | |
821 | */ | |
678aaf48 | 822 | jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode); |
d2a17637 MC |
823 | ei->i_reserved_data_blocks = 0; |
824 | ei->i_reserved_meta_blocks = 0; | |
825 | ei->i_allocated_meta_blocks = 0; | |
9d0be502 | 826 | ei->i_da_metadata_calc_len = 0; |
d2a17637 MC |
827 | ei->i_delalloc_reserved_flag = 0; |
828 | spin_lock_init(&(ei->i_block_reservation_lock)); | |
a9e7f447 DM |
829 | #ifdef CONFIG_QUOTA |
830 | ei->i_reserved_quota = 0; | |
831 | #endif | |
c7064ef1 | 832 | INIT_LIST_HEAD(&ei->i_completed_io_list); |
744692dc | 833 | spin_lock_init(&ei->i_completed_io_lock); |
8d5d02e6 | 834 | ei->cur_aio_dio = NULL; |
b436b9be JK |
835 | ei->i_sync_tid = 0; |
836 | ei->i_datasync_tid = 0; | |
f7ad6d2e | 837 | atomic_set(&ei->i_ioend_count, 0); |
0b8e58a1 | 838 | |
ac27a0ec DK |
839 | return &ei->vfs_inode; |
840 | } | |
841 | ||
7ff9c073 TT |
842 | static int ext4_drop_inode(struct inode *inode) |
843 | { | |
844 | int drop = generic_drop_inode(inode); | |
845 | ||
846 | trace_ext4_drop_inode(inode, drop); | |
847 | return drop; | |
848 | } | |
849 | ||
617ba13b | 850 | static void ext4_destroy_inode(struct inode *inode) |
ac27a0ec | 851 | { |
f7ad6d2e | 852 | ext4_ioend_wait(inode); |
9f7dd93d | 853 | if (!list_empty(&(EXT4_I(inode)->i_orphan))) { |
b31e1552 ES |
854 | ext4_msg(inode->i_sb, KERN_ERR, |
855 | "Inode %lu (%p): orphan list check failed!", | |
856 | inode->i_ino, EXT4_I(inode)); | |
9f7dd93d VA |
857 | print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, |
858 | EXT4_I(inode), sizeof(struct ext4_inode_info), | |
859 | true); | |
860 | dump_stack(); | |
861 | } | |
617ba13b | 862 | kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); |
ac27a0ec DK |
863 | } |
864 | ||
51cc5068 | 865 | static void init_once(void *foo) |
ac27a0ec | 866 | { |
617ba13b | 867 | struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; |
ac27a0ec | 868 | |
a35afb83 | 869 | INIT_LIST_HEAD(&ei->i_orphan); |
03010a33 | 870 | #ifdef CONFIG_EXT4_FS_XATTR |
a35afb83 | 871 | init_rwsem(&ei->xattr_sem); |
ac27a0ec | 872 | #endif |
0e855ac8 | 873 | init_rwsem(&ei->i_data_sem); |
a35afb83 | 874 | inode_init_once(&ei->vfs_inode); |
ac27a0ec DK |
875 | } |
876 | ||
877 | static int init_inodecache(void) | |
878 | { | |
617ba13b MC |
879 | ext4_inode_cachep = kmem_cache_create("ext4_inode_cache", |
880 | sizeof(struct ext4_inode_info), | |
ac27a0ec DK |
881 | 0, (SLAB_RECLAIM_ACCOUNT| |
882 | SLAB_MEM_SPREAD), | |
20c2df83 | 883 | init_once); |
617ba13b | 884 | if (ext4_inode_cachep == NULL) |
ac27a0ec DK |
885 | return -ENOMEM; |
886 | return 0; | |
887 | } | |
888 | ||
889 | static void destroy_inodecache(void) | |
890 | { | |
617ba13b | 891 | kmem_cache_destroy(ext4_inode_cachep); |
ac27a0ec DK |
892 | } |
893 | ||
0930fcc1 | 894 | void ext4_clear_inode(struct inode *inode) |
ac27a0ec | 895 | { |
0930fcc1 AV |
896 | invalidate_inode_buffers(inode); |
897 | end_writeback(inode); | |
9f754758 | 898 | dquot_drop(inode); |
c2ea3fde | 899 | ext4_discard_preallocations(inode); |
0390131b FM |
900 | if (EXT4_JOURNAL(inode)) |
901 | jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal, | |
678aaf48 | 902 | &EXT4_I(inode)->jinode); |
ac27a0ec DK |
903 | } |
904 | ||
2b2d6d01 TT |
905 | static inline void ext4_show_quota_options(struct seq_file *seq, |
906 | struct super_block *sb) | |
ac27a0ec DK |
907 | { |
908 | #if defined(CONFIG_QUOTA) | |
617ba13b | 909 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 910 | |
5a20bdfc JK |
911 | if (sbi->s_jquota_fmt) { |
912 | char *fmtname = ""; | |
913 | ||
914 | switch (sbi->s_jquota_fmt) { | |
915 | case QFMT_VFS_OLD: | |
916 | fmtname = "vfsold"; | |
917 | break; | |
918 | case QFMT_VFS_V0: | |
919 | fmtname = "vfsv0"; | |
920 | break; | |
921 | case QFMT_VFS_V1: | |
922 | fmtname = "vfsv1"; | |
923 | break; | |
924 | } | |
925 | seq_printf(seq, ",jqfmt=%s", fmtname); | |
926 | } | |
ac27a0ec DK |
927 | |
928 | if (sbi->s_qf_names[USRQUOTA]) | |
929 | seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]); | |
930 | ||
931 | if (sbi->s_qf_names[GRPQUOTA]) | |
932 | seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]); | |
933 | ||
482a7425 | 934 | if (test_opt(sb, USRQUOTA)) |
ac27a0ec DK |
935 | seq_puts(seq, ",usrquota"); |
936 | ||
482a7425 | 937 | if (test_opt(sb, GRPQUOTA)) |
ac27a0ec DK |
938 | seq_puts(seq, ",grpquota"); |
939 | #endif | |
940 | } | |
941 | ||
d9c9bef1 MS |
942 | /* |
943 | * Show an option if | |
944 | * - it's set to a non-default value OR | |
945 | * - if the per-sb default is different from the global default | |
946 | */ | |
617ba13b | 947 | static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs) |
ac27a0ec | 948 | { |
aa22df2c AK |
949 | int def_errors; |
950 | unsigned long def_mount_opts; | |
ac27a0ec | 951 | struct super_block *sb = vfs->mnt_sb; |
d9c9bef1 MS |
952 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
953 | struct ext4_super_block *es = sbi->s_es; | |
d9c9bef1 MS |
954 | |
955 | def_mount_opts = le32_to_cpu(es->s_default_mount_opts); | |
aa22df2c | 956 | def_errors = le16_to_cpu(es->s_errors); |
d9c9bef1 MS |
957 | |
958 | if (sbi->s_sb_block != 1) | |
959 | seq_printf(seq, ",sb=%llu", sbi->s_sb_block); | |
960 | if (test_opt(sb, MINIX_DF)) | |
961 | seq_puts(seq, ",minixdf"); | |
aa22df2c | 962 | if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS)) |
d9c9bef1 MS |
963 | seq_puts(seq, ",grpid"); |
964 | if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS)) | |
965 | seq_puts(seq, ",nogrpid"); | |
966 | if (sbi->s_resuid != EXT4_DEF_RESUID || | |
967 | le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) { | |
968 | seq_printf(seq, ",resuid=%u", sbi->s_resuid); | |
969 | } | |
970 | if (sbi->s_resgid != EXT4_DEF_RESGID || | |
971 | le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) { | |
972 | seq_printf(seq, ",resgid=%u", sbi->s_resgid); | |
973 | } | |
bb4f397a | 974 | if (test_opt(sb, ERRORS_RO)) { |
d9c9bef1 | 975 | if (def_errors == EXT4_ERRORS_PANIC || |
bb4f397a AK |
976 | def_errors == EXT4_ERRORS_CONTINUE) { |
977 | seq_puts(seq, ",errors=remount-ro"); | |
d9c9bef1 MS |
978 | } |
979 | } | |
aa22df2c | 980 | if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) |
bb4f397a | 981 | seq_puts(seq, ",errors=continue"); |
aa22df2c | 982 | if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) |
d9c9bef1 | 983 | seq_puts(seq, ",errors=panic"); |
aa22df2c | 984 | if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16)) |
d9c9bef1 | 985 | seq_puts(seq, ",nouid32"); |
aa22df2c | 986 | if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG)) |
d9c9bef1 MS |
987 | seq_puts(seq, ",debug"); |
988 | if (test_opt(sb, OLDALLOC)) | |
989 | seq_puts(seq, ",oldalloc"); | |
03010a33 | 990 | #ifdef CONFIG_EXT4_FS_XATTR |
aa22df2c AK |
991 | if (test_opt(sb, XATTR_USER) && |
992 | !(def_mount_opts & EXT4_DEFM_XATTR_USER)) | |
d9c9bef1 MS |
993 | seq_puts(seq, ",user_xattr"); |
994 | if (!test_opt(sb, XATTR_USER) && | |
995 | (def_mount_opts & EXT4_DEFM_XATTR_USER)) { | |
996 | seq_puts(seq, ",nouser_xattr"); | |
997 | } | |
998 | #endif | |
03010a33 | 999 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
aa22df2c | 1000 | if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL)) |
d9c9bef1 MS |
1001 | seq_puts(seq, ",acl"); |
1002 | if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL)) | |
1003 | seq_puts(seq, ",noacl"); | |
1004 | #endif | |
30773840 | 1005 | if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { |
d9c9bef1 MS |
1006 | seq_printf(seq, ",commit=%u", |
1007 | (unsigned) (sbi->s_commit_interval / HZ)); | |
1008 | } | |
30773840 TT |
1009 | if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) { |
1010 | seq_printf(seq, ",min_batch_time=%u", | |
1011 | (unsigned) sbi->s_min_batch_time); | |
1012 | } | |
1013 | if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) { | |
1014 | seq_printf(seq, ",max_batch_time=%u", | |
1015 | (unsigned) sbi->s_min_batch_time); | |
1016 | } | |
1017 | ||
571640ca ES |
1018 | /* |
1019 | * We're changing the default of barrier mount option, so | |
1020 | * let's always display its mount state so it's clear what its | |
1021 | * status is. | |
1022 | */ | |
1023 | seq_puts(seq, ",barrier="); | |
1024 | seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0"); | |
cd0b6a39 TT |
1025 | if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) |
1026 | seq_puts(seq, ",journal_async_commit"); | |
39a4bade JK |
1027 | else if (test_opt(sb, JOURNAL_CHECKSUM)) |
1028 | seq_puts(seq, ",journal_checksum"); | |
25ec56b5 JNC |
1029 | if (test_opt(sb, I_VERSION)) |
1030 | seq_puts(seq, ",i_version"); | |
8b67f04a TT |
1031 | if (!test_opt(sb, DELALLOC) && |
1032 | !(def_mount_opts & EXT4_DEFM_NODELALLOC)) | |
dd919b98 AK |
1033 | seq_puts(seq, ",nodelalloc"); |
1034 | ||
1449032b TT |
1035 | if (test_opt(sb, MBLK_IO_SUBMIT)) |
1036 | seq_puts(seq, ",mblk_io_submit"); | |
cb45bbe4 MS |
1037 | if (sbi->s_stripe) |
1038 | seq_printf(seq, ",stripe=%lu", sbi->s_stripe); | |
aa22df2c AK |
1039 | /* |
1040 | * journal mode get enabled in different ways | |
1041 | * So just print the value even if we didn't specify it | |
1042 | */ | |
617ba13b | 1043 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) |
ac27a0ec | 1044 | seq_puts(seq, ",data=journal"); |
617ba13b | 1045 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) |
ac27a0ec | 1046 | seq_puts(seq, ",data=ordered"); |
617ba13b | 1047 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) |
ac27a0ec DK |
1048 | seq_puts(seq, ",data=writeback"); |
1049 | ||
240799cd TT |
1050 | if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) |
1051 | seq_printf(seq, ",inode_readahead_blks=%u", | |
1052 | sbi->s_inode_readahead_blks); | |
1053 | ||
5bf5683a HK |
1054 | if (test_opt(sb, DATA_ERR_ABORT)) |
1055 | seq_puts(seq, ",data_err=abort"); | |
1056 | ||
afd4672d | 1057 | if (test_opt(sb, NO_AUTO_DA_ALLOC)) |
06705bff | 1058 | seq_puts(seq, ",noauto_da_alloc"); |
afd4672d | 1059 | |
8b67f04a | 1060 | if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD)) |
5328e635 ES |
1061 | seq_puts(seq, ",discard"); |
1062 | ||
e3bb52ae ES |
1063 | if (test_opt(sb, NOLOAD)) |
1064 | seq_puts(seq, ",norecovery"); | |
1065 | ||
744692dc JZ |
1066 | if (test_opt(sb, DIOREAD_NOLOCK)) |
1067 | seq_puts(seq, ",dioread_nolock"); | |
1068 | ||
8b67f04a TT |
1069 | if (test_opt(sb, BLOCK_VALIDITY) && |
1070 | !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)) | |
1071 | seq_puts(seq, ",block_validity"); | |
1072 | ||
bfff6873 LC |
1073 | if (!test_opt(sb, INIT_INODE_TABLE)) |
1074 | seq_puts(seq, ",noinit_inode_table"); | |
1075 | else if (sbi->s_li_wait_mult) | |
1076 | seq_printf(seq, ",init_inode_table=%u", | |
1077 | (unsigned) sbi->s_li_wait_mult); | |
1078 | ||
617ba13b | 1079 | ext4_show_quota_options(seq, sb); |
0b8e58a1 | 1080 | |
ac27a0ec DK |
1081 | return 0; |
1082 | } | |
1083 | ||
1b961ac0 | 1084 | static struct inode *ext4_nfs_get_inode(struct super_block *sb, |
0b8e58a1 | 1085 | u64 ino, u32 generation) |
ac27a0ec | 1086 | { |
ac27a0ec | 1087 | struct inode *inode; |
ac27a0ec | 1088 | |
617ba13b | 1089 | if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO) |
ac27a0ec | 1090 | return ERR_PTR(-ESTALE); |
617ba13b | 1091 | if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count)) |
ac27a0ec DK |
1092 | return ERR_PTR(-ESTALE); |
1093 | ||
1094 | /* iget isn't really right if the inode is currently unallocated!! | |
1095 | * | |
617ba13b | 1096 | * ext4_read_inode will return a bad_inode if the inode had been |
ac27a0ec DK |
1097 | * deleted, so we should be safe. |
1098 | * | |
1099 | * Currently we don't know the generation for parent directory, so | |
1100 | * a generation of 0 means "accept any" | |
1101 | */ | |
1d1fe1ee DH |
1102 | inode = ext4_iget(sb, ino); |
1103 | if (IS_ERR(inode)) | |
1104 | return ERR_CAST(inode); | |
1105 | if (generation && inode->i_generation != generation) { | |
ac27a0ec DK |
1106 | iput(inode); |
1107 | return ERR_PTR(-ESTALE); | |
1108 | } | |
1b961ac0 CH |
1109 | |
1110 | return inode; | |
1111 | } | |
1112 | ||
1113 | static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, | |
0b8e58a1 | 1114 | int fh_len, int fh_type) |
1b961ac0 CH |
1115 | { |
1116 | return generic_fh_to_dentry(sb, fid, fh_len, fh_type, | |
1117 | ext4_nfs_get_inode); | |
1118 | } | |
1119 | ||
1120 | static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, | |
0b8e58a1 | 1121 | int fh_len, int fh_type) |
1b961ac0 CH |
1122 | { |
1123 | return generic_fh_to_parent(sb, fid, fh_len, fh_type, | |
1124 | ext4_nfs_get_inode); | |
ac27a0ec DK |
1125 | } |
1126 | ||
c39a7f84 TO |
1127 | /* |
1128 | * Try to release metadata pages (indirect blocks, directories) which are | |
1129 | * mapped via the block device. Since these pages could have journal heads | |
1130 | * which would prevent try_to_free_buffers() from freeing them, we must use | |
1131 | * jbd2 layer's try_to_free_buffers() function to release them. | |
1132 | */ | |
0b8e58a1 AD |
1133 | static int bdev_try_to_free_page(struct super_block *sb, struct page *page, |
1134 | gfp_t wait) | |
c39a7f84 TO |
1135 | { |
1136 | journal_t *journal = EXT4_SB(sb)->s_journal; | |
1137 | ||
1138 | WARN_ON(PageChecked(page)); | |
1139 | if (!page_has_buffers(page)) | |
1140 | return 0; | |
1141 | if (journal) | |
1142 | return jbd2_journal_try_to_free_buffers(journal, page, | |
1143 | wait & ~__GFP_WAIT); | |
1144 | return try_to_free_buffers(page); | |
1145 | } | |
1146 | ||
ac27a0ec | 1147 | #ifdef CONFIG_QUOTA |
af5bc92d | 1148 | #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group") |
2b2d6d01 | 1149 | #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA)) |
ac27a0ec | 1150 | |
617ba13b MC |
1151 | static int ext4_write_dquot(struct dquot *dquot); |
1152 | static int ext4_acquire_dquot(struct dquot *dquot); | |
1153 | static int ext4_release_dquot(struct dquot *dquot); | |
1154 | static int ext4_mark_dquot_dirty(struct dquot *dquot); | |
1155 | static int ext4_write_info(struct super_block *sb, int type); | |
6f28e087 | 1156 | static int ext4_quota_on(struct super_block *sb, int type, int format_id, |
307ae18a | 1157 | char *path); |
ca0e05e4 | 1158 | static int ext4_quota_off(struct super_block *sb, int type); |
617ba13b MC |
1159 | static int ext4_quota_on_mount(struct super_block *sb, int type); |
1160 | static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, | |
ac27a0ec | 1161 | size_t len, loff_t off); |
617ba13b | 1162 | static ssize_t ext4_quota_write(struct super_block *sb, int type, |
ac27a0ec DK |
1163 | const char *data, size_t len, loff_t off); |
1164 | ||
61e225dc | 1165 | static const struct dquot_operations ext4_quota_operations = { |
a9e7f447 | 1166 | #ifdef CONFIG_QUOTA |
60e58e0f | 1167 | .get_reserved_space = ext4_get_reserved_space, |
a9e7f447 | 1168 | #endif |
617ba13b MC |
1169 | .write_dquot = ext4_write_dquot, |
1170 | .acquire_dquot = ext4_acquire_dquot, | |
1171 | .release_dquot = ext4_release_dquot, | |
1172 | .mark_dirty = ext4_mark_dquot_dirty, | |
a5b5ee32 JK |
1173 | .write_info = ext4_write_info, |
1174 | .alloc_dquot = dquot_alloc, | |
1175 | .destroy_dquot = dquot_destroy, | |
ac27a0ec DK |
1176 | }; |
1177 | ||
0d54b217 | 1178 | static const struct quotactl_ops ext4_qctl_operations = { |
617ba13b | 1179 | .quota_on = ext4_quota_on, |
ca0e05e4 | 1180 | .quota_off = ext4_quota_off, |
287a8095 CH |
1181 | .quota_sync = dquot_quota_sync, |
1182 | .get_info = dquot_get_dqinfo, | |
1183 | .set_info = dquot_set_dqinfo, | |
1184 | .get_dqblk = dquot_get_dqblk, | |
1185 | .set_dqblk = dquot_set_dqblk | |
ac27a0ec DK |
1186 | }; |
1187 | #endif | |
1188 | ||
ee9b6d61 | 1189 | static const struct super_operations ext4_sops = { |
617ba13b MC |
1190 | .alloc_inode = ext4_alloc_inode, |
1191 | .destroy_inode = ext4_destroy_inode, | |
617ba13b MC |
1192 | .write_inode = ext4_write_inode, |
1193 | .dirty_inode = ext4_dirty_inode, | |
7ff9c073 | 1194 | .drop_inode = ext4_drop_inode, |
0930fcc1 | 1195 | .evict_inode = ext4_evict_inode, |
617ba13b | 1196 | .put_super = ext4_put_super, |
617ba13b | 1197 | .sync_fs = ext4_sync_fs, |
c4be0c1d TS |
1198 | .freeze_fs = ext4_freeze, |
1199 | .unfreeze_fs = ext4_unfreeze, | |
617ba13b MC |
1200 | .statfs = ext4_statfs, |
1201 | .remount_fs = ext4_remount, | |
617ba13b | 1202 | .show_options = ext4_show_options, |
ac27a0ec | 1203 | #ifdef CONFIG_QUOTA |
617ba13b MC |
1204 | .quota_read = ext4_quota_read, |
1205 | .quota_write = ext4_quota_write, | |
ac27a0ec | 1206 | #endif |
c39a7f84 | 1207 | .bdev_try_to_free_page = bdev_try_to_free_page, |
ac27a0ec DK |
1208 | }; |
1209 | ||
9ca92389 TT |
1210 | static const struct super_operations ext4_nojournal_sops = { |
1211 | .alloc_inode = ext4_alloc_inode, | |
1212 | .destroy_inode = ext4_destroy_inode, | |
1213 | .write_inode = ext4_write_inode, | |
1214 | .dirty_inode = ext4_dirty_inode, | |
7ff9c073 | 1215 | .drop_inode = ext4_drop_inode, |
0930fcc1 | 1216 | .evict_inode = ext4_evict_inode, |
9ca92389 TT |
1217 | .write_super = ext4_write_super, |
1218 | .put_super = ext4_put_super, | |
1219 | .statfs = ext4_statfs, | |
1220 | .remount_fs = ext4_remount, | |
9ca92389 TT |
1221 | .show_options = ext4_show_options, |
1222 | #ifdef CONFIG_QUOTA | |
1223 | .quota_read = ext4_quota_read, | |
1224 | .quota_write = ext4_quota_write, | |
1225 | #endif | |
1226 | .bdev_try_to_free_page = bdev_try_to_free_page, | |
1227 | }; | |
1228 | ||
39655164 | 1229 | static const struct export_operations ext4_export_ops = { |
1b961ac0 CH |
1230 | .fh_to_dentry = ext4_fh_to_dentry, |
1231 | .fh_to_parent = ext4_fh_to_parent, | |
617ba13b | 1232 | .get_parent = ext4_get_parent, |
ac27a0ec DK |
1233 | }; |
1234 | ||
1235 | enum { | |
1236 | Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, | |
1237 | Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, | |
01436ef2 | 1238 | Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov, |
ac27a0ec | 1239 | Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, |
06705bff | 1240 | Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh, |
30773840 | 1241 | Opt_commit, Opt_min_batch_time, Opt_max_batch_time, |
c3191067 | 1242 | Opt_journal_update, Opt_journal_dev, |
818d276c | 1243 | Opt_journal_checksum, Opt_journal_async_commit, |
ac27a0ec | 1244 | Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, |
296c355c | 1245 | Opt_data_err_abort, Opt_data_err_ignore, |
ac27a0ec | 1246 | Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, |
5a20bdfc JK |
1247 | Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, |
1248 | Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, | |
1249 | Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version, | |
1449032b TT |
1250 | Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit, |
1251 | Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, | |
5328e635 | 1252 | Opt_inode_readahead_blks, Opt_journal_ioprio, |
744692dc | 1253 | Opt_dioread_nolock, Opt_dioread_lock, |
5328e635 | 1254 | Opt_discard, Opt_nodiscard, |
bfff6873 | 1255 | Opt_init_inode_table, Opt_noinit_inode_table, |
ac27a0ec DK |
1256 | }; |
1257 | ||
a447c093 | 1258 | static const match_table_t tokens = { |
ac27a0ec DK |
1259 | {Opt_bsd_df, "bsddf"}, |
1260 | {Opt_minix_df, "minixdf"}, | |
1261 | {Opt_grpid, "grpid"}, | |
1262 | {Opt_grpid, "bsdgroups"}, | |
1263 | {Opt_nogrpid, "nogrpid"}, | |
1264 | {Opt_nogrpid, "sysvgroups"}, | |
1265 | {Opt_resgid, "resgid=%u"}, | |
1266 | {Opt_resuid, "resuid=%u"}, | |
1267 | {Opt_sb, "sb=%u"}, | |
1268 | {Opt_err_cont, "errors=continue"}, | |
1269 | {Opt_err_panic, "errors=panic"}, | |
1270 | {Opt_err_ro, "errors=remount-ro"}, | |
1271 | {Opt_nouid32, "nouid32"}, | |
ac27a0ec DK |
1272 | {Opt_debug, "debug"}, |
1273 | {Opt_oldalloc, "oldalloc"}, | |
1274 | {Opt_orlov, "orlov"}, | |
1275 | {Opt_user_xattr, "user_xattr"}, | |
1276 | {Opt_nouser_xattr, "nouser_xattr"}, | |
1277 | {Opt_acl, "acl"}, | |
1278 | {Opt_noacl, "noacl"}, | |
ac27a0ec | 1279 | {Opt_noload, "noload"}, |
e3bb52ae | 1280 | {Opt_noload, "norecovery"}, |
ac27a0ec DK |
1281 | {Opt_nobh, "nobh"}, |
1282 | {Opt_bh, "bh"}, | |
1283 | {Opt_commit, "commit=%u"}, | |
30773840 TT |
1284 | {Opt_min_batch_time, "min_batch_time=%u"}, |
1285 | {Opt_max_batch_time, "max_batch_time=%u"}, | |
ac27a0ec | 1286 | {Opt_journal_update, "journal=update"}, |
ac27a0ec | 1287 | {Opt_journal_dev, "journal_dev=%u"}, |
818d276c GS |
1288 | {Opt_journal_checksum, "journal_checksum"}, |
1289 | {Opt_journal_async_commit, "journal_async_commit"}, | |
ac27a0ec DK |
1290 | {Opt_abort, "abort"}, |
1291 | {Opt_data_journal, "data=journal"}, | |
1292 | {Opt_data_ordered, "data=ordered"}, | |
1293 | {Opt_data_writeback, "data=writeback"}, | |
5bf5683a HK |
1294 | {Opt_data_err_abort, "data_err=abort"}, |
1295 | {Opt_data_err_ignore, "data_err=ignore"}, | |
ac27a0ec DK |
1296 | {Opt_offusrjquota, "usrjquota="}, |
1297 | {Opt_usrjquota, "usrjquota=%s"}, | |
1298 | {Opt_offgrpjquota, "grpjquota="}, | |
1299 | {Opt_grpjquota, "grpjquota=%s"}, | |
1300 | {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, | |
1301 | {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, | |
5a20bdfc | 1302 | {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, |
ac27a0ec DK |
1303 | {Opt_grpquota, "grpquota"}, |
1304 | {Opt_noquota, "noquota"}, | |
1305 | {Opt_quota, "quota"}, | |
1306 | {Opt_usrquota, "usrquota"}, | |
1307 | {Opt_barrier, "barrier=%u"}, | |
06705bff TT |
1308 | {Opt_barrier, "barrier"}, |
1309 | {Opt_nobarrier, "nobarrier"}, | |
25ec56b5 | 1310 | {Opt_i_version, "i_version"}, |
c9de560d | 1311 | {Opt_stripe, "stripe=%u"}, |
ac27a0ec | 1312 | {Opt_resize, "resize"}, |
64769240 | 1313 | {Opt_delalloc, "delalloc"}, |
dd919b98 | 1314 | {Opt_nodelalloc, "nodelalloc"}, |
1449032b TT |
1315 | {Opt_mblk_io_submit, "mblk_io_submit"}, |
1316 | {Opt_nomblk_io_submit, "nomblk_io_submit"}, | |
6fd058f7 TT |
1317 | {Opt_block_validity, "block_validity"}, |
1318 | {Opt_noblock_validity, "noblock_validity"}, | |
240799cd | 1319 | {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, |
b3881f74 | 1320 | {Opt_journal_ioprio, "journal_ioprio=%u"}, |
afd4672d | 1321 | {Opt_auto_da_alloc, "auto_da_alloc=%u"}, |
06705bff TT |
1322 | {Opt_auto_da_alloc, "auto_da_alloc"}, |
1323 | {Opt_noauto_da_alloc, "noauto_da_alloc"}, | |
744692dc JZ |
1324 | {Opt_dioread_nolock, "dioread_nolock"}, |
1325 | {Opt_dioread_lock, "dioread_lock"}, | |
5328e635 ES |
1326 | {Opt_discard, "discard"}, |
1327 | {Opt_nodiscard, "nodiscard"}, | |
bfff6873 LC |
1328 | {Opt_init_inode_table, "init_itable=%u"}, |
1329 | {Opt_init_inode_table, "init_itable"}, | |
1330 | {Opt_noinit_inode_table, "noinit_itable"}, | |
f3f12faa | 1331 | {Opt_err, NULL}, |
ac27a0ec DK |
1332 | }; |
1333 | ||
617ba13b | 1334 | static ext4_fsblk_t get_sb_block(void **data) |
ac27a0ec | 1335 | { |
617ba13b | 1336 | ext4_fsblk_t sb_block; |
ac27a0ec DK |
1337 | char *options = (char *) *data; |
1338 | ||
1339 | if (!options || strncmp(options, "sb=", 3) != 0) | |
1340 | return 1; /* Default location */ | |
0b8e58a1 | 1341 | |
ac27a0ec | 1342 | options += 3; |
0b8e58a1 | 1343 | /* TODO: use simple_strtoll with >32bit ext4 */ |
ac27a0ec DK |
1344 | sb_block = simple_strtoul(options, &options, 0); |
1345 | if (*options && *options != ',') { | |
4776004f | 1346 | printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", |
ac27a0ec DK |
1347 | (char *) *data); |
1348 | return 1; | |
1349 | } | |
1350 | if (*options == ',') | |
1351 | options++; | |
1352 | *data = (void *) options; | |
0b8e58a1 | 1353 | |
ac27a0ec DK |
1354 | return sb_block; |
1355 | } | |
1356 | ||
b3881f74 | 1357 | #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) |
437ca0fd DM |
1358 | static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n" |
1359 | "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; | |
b3881f74 | 1360 | |
56c50f11 DM |
1361 | #ifdef CONFIG_QUOTA |
1362 | static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) | |
1363 | { | |
1364 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1365 | char *qname; | |
1366 | ||
1367 | if (sb_any_quota_loaded(sb) && | |
1368 | !sbi->s_qf_names[qtype]) { | |
1369 | ext4_msg(sb, KERN_ERR, | |
1370 | "Cannot change journaled " | |
1371 | "quota options when quota turned on"); | |
1372 | return 0; | |
1373 | } | |
1374 | qname = match_strdup(args); | |
1375 | if (!qname) { | |
1376 | ext4_msg(sb, KERN_ERR, | |
1377 | "Not enough memory for storing quotafile name"); | |
1378 | return 0; | |
1379 | } | |
1380 | if (sbi->s_qf_names[qtype] && | |
1381 | strcmp(sbi->s_qf_names[qtype], qname)) { | |
1382 | ext4_msg(sb, KERN_ERR, | |
1383 | "%s quota file already specified", QTYPE2NAME(qtype)); | |
1384 | kfree(qname); | |
1385 | return 0; | |
1386 | } | |
1387 | sbi->s_qf_names[qtype] = qname; | |
1388 | if (strchr(sbi->s_qf_names[qtype], '/')) { | |
1389 | ext4_msg(sb, KERN_ERR, | |
1390 | "quotafile must be on filesystem root"); | |
1391 | kfree(sbi->s_qf_names[qtype]); | |
1392 | sbi->s_qf_names[qtype] = NULL; | |
1393 | return 0; | |
1394 | } | |
fd8c37ec | 1395 | set_opt(sb, QUOTA); |
56c50f11 DM |
1396 | return 1; |
1397 | } | |
1398 | ||
1399 | static int clear_qf_name(struct super_block *sb, int qtype) | |
1400 | { | |
1401 | ||
1402 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1403 | ||
1404 | if (sb_any_quota_loaded(sb) && | |
1405 | sbi->s_qf_names[qtype]) { | |
1406 | ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" | |
1407 | " when quota turned on"); | |
1408 | return 0; | |
1409 | } | |
1410 | /* | |
1411 | * The space will be released later when all options are confirmed | |
1412 | * to be correct | |
1413 | */ | |
1414 | sbi->s_qf_names[qtype] = NULL; | |
1415 | return 1; | |
1416 | } | |
1417 | #endif | |
1418 | ||
2b2d6d01 | 1419 | static int parse_options(char *options, struct super_block *sb, |
c3191067 | 1420 | unsigned long *journal_devnum, |
b3881f74 | 1421 | unsigned int *journal_ioprio, |
2b2d6d01 | 1422 | ext4_fsblk_t *n_blocks_count, int is_remount) |
ac27a0ec | 1423 | { |
617ba13b | 1424 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
2b2d6d01 | 1425 | char *p; |
ac27a0ec DK |
1426 | substring_t args[MAX_OPT_ARGS]; |
1427 | int data_opt = 0; | |
1428 | int option; | |
1429 | #ifdef CONFIG_QUOTA | |
56c50f11 | 1430 | int qfmt; |
ac27a0ec DK |
1431 | #endif |
1432 | ||
1433 | if (!options) | |
1434 | return 1; | |
1435 | ||
2b2d6d01 | 1436 | while ((p = strsep(&options, ",")) != NULL) { |
ac27a0ec DK |
1437 | int token; |
1438 | if (!*p) | |
1439 | continue; | |
1440 | ||
15121c18 ES |
1441 | /* |
1442 | * Initialize args struct so we know whether arg was | |
1443 | * found; some options take optional arguments. | |
1444 | */ | |
1445 | args[0].to = args[0].from = 0; | |
ac27a0ec DK |
1446 | token = match_token(p, tokens, args); |
1447 | switch (token) { | |
1448 | case Opt_bsd_df: | |
437ca0fd | 1449 | ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); |
fd8c37ec | 1450 | clear_opt(sb, MINIX_DF); |
ac27a0ec DK |
1451 | break; |
1452 | case Opt_minix_df: | |
437ca0fd | 1453 | ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); |
fd8c37ec | 1454 | set_opt(sb, MINIX_DF); |
437ca0fd | 1455 | |
ac27a0ec DK |
1456 | break; |
1457 | case Opt_grpid: | |
437ca0fd | 1458 | ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); |
fd8c37ec | 1459 | set_opt(sb, GRPID); |
437ca0fd | 1460 | |
ac27a0ec DK |
1461 | break; |
1462 | case Opt_nogrpid: | |
437ca0fd | 1463 | ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38"); |
fd8c37ec | 1464 | clear_opt(sb, GRPID); |
437ca0fd | 1465 | |
ac27a0ec DK |
1466 | break; |
1467 | case Opt_resuid: | |
1468 | if (match_int(&args[0], &option)) | |
1469 | return 0; | |
1470 | sbi->s_resuid = option; | |
1471 | break; | |
1472 | case Opt_resgid: | |
1473 | if (match_int(&args[0], &option)) | |
1474 | return 0; | |
1475 | sbi->s_resgid = option; | |
1476 | break; | |
1477 | case Opt_sb: | |
1478 | /* handled by get_sb_block() instead of here */ | |
1479 | /* *sb_block = match_int(&args[0]); */ | |
1480 | break; | |
1481 | case Opt_err_panic: | |
fd8c37ec TT |
1482 | clear_opt(sb, ERRORS_CONT); |
1483 | clear_opt(sb, ERRORS_RO); | |
1484 | set_opt(sb, ERRORS_PANIC); | |
ac27a0ec DK |
1485 | break; |
1486 | case Opt_err_ro: | |
fd8c37ec TT |
1487 | clear_opt(sb, ERRORS_CONT); |
1488 | clear_opt(sb, ERRORS_PANIC); | |
1489 | set_opt(sb, ERRORS_RO); | |
ac27a0ec DK |
1490 | break; |
1491 | case Opt_err_cont: | |
fd8c37ec TT |
1492 | clear_opt(sb, ERRORS_RO); |
1493 | clear_opt(sb, ERRORS_PANIC); | |
1494 | set_opt(sb, ERRORS_CONT); | |
ac27a0ec DK |
1495 | break; |
1496 | case Opt_nouid32: | |
fd8c37ec | 1497 | set_opt(sb, NO_UID32); |
ac27a0ec | 1498 | break; |
ac27a0ec | 1499 | case Opt_debug: |
fd8c37ec | 1500 | set_opt(sb, DEBUG); |
ac27a0ec DK |
1501 | break; |
1502 | case Opt_oldalloc: | |
fd8c37ec | 1503 | set_opt(sb, OLDALLOC); |
ac27a0ec DK |
1504 | break; |
1505 | case Opt_orlov: | |
fd8c37ec | 1506 | clear_opt(sb, OLDALLOC); |
ac27a0ec | 1507 | break; |
03010a33 | 1508 | #ifdef CONFIG_EXT4_FS_XATTR |
ac27a0ec | 1509 | case Opt_user_xattr: |
fd8c37ec | 1510 | set_opt(sb, XATTR_USER); |
ac27a0ec DK |
1511 | break; |
1512 | case Opt_nouser_xattr: | |
fd8c37ec | 1513 | clear_opt(sb, XATTR_USER); |
ac27a0ec DK |
1514 | break; |
1515 | #else | |
1516 | case Opt_user_xattr: | |
1517 | case Opt_nouser_xattr: | |
b31e1552 | 1518 | ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported"); |
ac27a0ec DK |
1519 | break; |
1520 | #endif | |
03010a33 | 1521 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
ac27a0ec | 1522 | case Opt_acl: |
fd8c37ec | 1523 | set_opt(sb, POSIX_ACL); |
ac27a0ec DK |
1524 | break; |
1525 | case Opt_noacl: | |
fd8c37ec | 1526 | clear_opt(sb, POSIX_ACL); |
ac27a0ec DK |
1527 | break; |
1528 | #else | |
1529 | case Opt_acl: | |
1530 | case Opt_noacl: | |
b31e1552 | 1531 | ext4_msg(sb, KERN_ERR, "(no)acl options not supported"); |
ac27a0ec DK |
1532 | break; |
1533 | #endif | |
ac27a0ec DK |
1534 | case Opt_journal_update: |
1535 | /* @@@ FIXME */ | |
1536 | /* Eventually we will want to be able to create | |
1537 | a journal file here. For now, only allow the | |
1538 | user to specify an existing inode to be the | |
1539 | journal file. */ | |
1540 | if (is_remount) { | |
b31e1552 ES |
1541 | ext4_msg(sb, KERN_ERR, |
1542 | "Cannot specify journal on remount"); | |
ac27a0ec DK |
1543 | return 0; |
1544 | } | |
fd8c37ec | 1545 | set_opt(sb, UPDATE_JOURNAL); |
ac27a0ec | 1546 | break; |
ac27a0ec DK |
1547 | case Opt_journal_dev: |
1548 | if (is_remount) { | |
b31e1552 ES |
1549 | ext4_msg(sb, KERN_ERR, |
1550 | "Cannot specify journal on remount"); | |
ac27a0ec DK |
1551 | return 0; |
1552 | } | |
1553 | if (match_int(&args[0], &option)) | |
1554 | return 0; | |
1555 | *journal_devnum = option; | |
1556 | break; | |
818d276c | 1557 | case Opt_journal_checksum: |
fd8c37ec | 1558 | set_opt(sb, JOURNAL_CHECKSUM); |
d4da6c9c | 1559 | break; |
818d276c | 1560 | case Opt_journal_async_commit: |
fd8c37ec TT |
1561 | set_opt(sb, JOURNAL_ASYNC_COMMIT); |
1562 | set_opt(sb, JOURNAL_CHECKSUM); | |
818d276c | 1563 | break; |
ac27a0ec | 1564 | case Opt_noload: |
fd8c37ec | 1565 | set_opt(sb, NOLOAD); |
ac27a0ec DK |
1566 | break; |
1567 | case Opt_commit: | |
1568 | if (match_int(&args[0], &option)) | |
1569 | return 0; | |
1570 | if (option < 0) | |
1571 | return 0; | |
1572 | if (option == 0) | |
cd02ff0b | 1573 | option = JBD2_DEFAULT_MAX_COMMIT_AGE; |
ac27a0ec DK |
1574 | sbi->s_commit_interval = HZ * option; |
1575 | break; | |
30773840 TT |
1576 | case Opt_max_batch_time: |
1577 | if (match_int(&args[0], &option)) | |
1578 | return 0; | |
1579 | if (option < 0) | |
1580 | return 0; | |
1581 | if (option == 0) | |
1582 | option = EXT4_DEF_MAX_BATCH_TIME; | |
1583 | sbi->s_max_batch_time = option; | |
1584 | break; | |
1585 | case Opt_min_batch_time: | |
1586 | if (match_int(&args[0], &option)) | |
1587 | return 0; | |
1588 | if (option < 0) | |
1589 | return 0; | |
1590 | sbi->s_min_batch_time = option; | |
1591 | break; | |
ac27a0ec | 1592 | case Opt_data_journal: |
617ba13b | 1593 | data_opt = EXT4_MOUNT_JOURNAL_DATA; |
ac27a0ec DK |
1594 | goto datacheck; |
1595 | case Opt_data_ordered: | |
617ba13b | 1596 | data_opt = EXT4_MOUNT_ORDERED_DATA; |
ac27a0ec DK |
1597 | goto datacheck; |
1598 | case Opt_data_writeback: | |
617ba13b | 1599 | data_opt = EXT4_MOUNT_WRITEBACK_DATA; |
ac27a0ec DK |
1600 | datacheck: |
1601 | if (is_remount) { | |
482a7425 | 1602 | if (test_opt(sb, DATA_FLAGS) != data_opt) { |
b31e1552 ES |
1603 | ext4_msg(sb, KERN_ERR, |
1604 | "Cannot change data mode on remount"); | |
ac27a0ec DK |
1605 | return 0; |
1606 | } | |
1607 | } else { | |
fd8c37ec | 1608 | clear_opt(sb, DATA_FLAGS); |
ac27a0ec DK |
1609 | sbi->s_mount_opt |= data_opt; |
1610 | } | |
1611 | break; | |
5bf5683a | 1612 | case Opt_data_err_abort: |
fd8c37ec | 1613 | set_opt(sb, DATA_ERR_ABORT); |
5bf5683a HK |
1614 | break; |
1615 | case Opt_data_err_ignore: | |
fd8c37ec | 1616 | clear_opt(sb, DATA_ERR_ABORT); |
5bf5683a | 1617 | break; |
ac27a0ec DK |
1618 | #ifdef CONFIG_QUOTA |
1619 | case Opt_usrjquota: | |
56c50f11 | 1620 | if (!set_qf_name(sb, USRQUOTA, &args[0])) |
ac27a0ec | 1621 | return 0; |
56c50f11 DM |
1622 | break; |
1623 | case Opt_grpjquota: | |
1624 | if (!set_qf_name(sb, GRPQUOTA, &args[0])) | |
ac27a0ec | 1625 | return 0; |
ac27a0ec DK |
1626 | break; |
1627 | case Opt_offusrjquota: | |
56c50f11 DM |
1628 | if (!clear_qf_name(sb, USRQUOTA)) |
1629 | return 0; | |
1630 | break; | |
ac27a0ec | 1631 | case Opt_offgrpjquota: |
56c50f11 | 1632 | if (!clear_qf_name(sb, GRPQUOTA)) |
ac27a0ec | 1633 | return 0; |
ac27a0ec | 1634 | break; |
56c50f11 | 1635 | |
ac27a0ec | 1636 | case Opt_jqfmt_vfsold: |
dfc5d03f JK |
1637 | qfmt = QFMT_VFS_OLD; |
1638 | goto set_qf_format; | |
ac27a0ec | 1639 | case Opt_jqfmt_vfsv0: |
dfc5d03f | 1640 | qfmt = QFMT_VFS_V0; |
5a20bdfc JK |
1641 | goto set_qf_format; |
1642 | case Opt_jqfmt_vfsv1: | |
1643 | qfmt = QFMT_VFS_V1; | |
dfc5d03f | 1644 | set_qf_format: |
17bd13b3 | 1645 | if (sb_any_quota_loaded(sb) && |
dfc5d03f | 1646 | sbi->s_jquota_fmt != qfmt) { |
b31e1552 | 1647 | ext4_msg(sb, KERN_ERR, "Cannot change " |
dfc5d03f | 1648 | "journaled quota options when " |
b31e1552 | 1649 | "quota turned on"); |
dfc5d03f JK |
1650 | return 0; |
1651 | } | |
1652 | sbi->s_jquota_fmt = qfmt; | |
ac27a0ec DK |
1653 | break; |
1654 | case Opt_quota: | |
1655 | case Opt_usrquota: | |
fd8c37ec TT |
1656 | set_opt(sb, QUOTA); |
1657 | set_opt(sb, USRQUOTA); | |
ac27a0ec DK |
1658 | break; |
1659 | case Opt_grpquota: | |
fd8c37ec TT |
1660 | set_opt(sb, QUOTA); |
1661 | set_opt(sb, GRPQUOTA); | |
ac27a0ec DK |
1662 | break; |
1663 | case Opt_noquota: | |
17bd13b3 | 1664 | if (sb_any_quota_loaded(sb)) { |
b31e1552 ES |
1665 | ext4_msg(sb, KERN_ERR, "Cannot change quota " |
1666 | "options when quota turned on"); | |
ac27a0ec DK |
1667 | return 0; |
1668 | } | |
fd8c37ec TT |
1669 | clear_opt(sb, QUOTA); |
1670 | clear_opt(sb, USRQUOTA); | |
1671 | clear_opt(sb, GRPQUOTA); | |
ac27a0ec DK |
1672 | break; |
1673 | #else | |
1674 | case Opt_quota: | |
1675 | case Opt_usrquota: | |
1676 | case Opt_grpquota: | |
b31e1552 ES |
1677 | ext4_msg(sb, KERN_ERR, |
1678 | "quota options not supported"); | |
cd59e7b9 | 1679 | break; |
ac27a0ec DK |
1680 | case Opt_usrjquota: |
1681 | case Opt_grpjquota: | |
1682 | case Opt_offusrjquota: | |
1683 | case Opt_offgrpjquota: | |
1684 | case Opt_jqfmt_vfsold: | |
1685 | case Opt_jqfmt_vfsv0: | |
5a20bdfc | 1686 | case Opt_jqfmt_vfsv1: |
b31e1552 ES |
1687 | ext4_msg(sb, KERN_ERR, |
1688 | "journaled quota options not supported"); | |
ac27a0ec DK |
1689 | break; |
1690 | case Opt_noquota: | |
1691 | break; | |
1692 | #endif | |
1693 | case Opt_abort: | |
4ab2f15b | 1694 | sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; |
ac27a0ec | 1695 | break; |
06705bff | 1696 | case Opt_nobarrier: |
fd8c37ec | 1697 | clear_opt(sb, BARRIER); |
06705bff | 1698 | break; |
ac27a0ec | 1699 | case Opt_barrier: |
15121c18 ES |
1700 | if (args[0].from) { |
1701 | if (match_int(&args[0], &option)) | |
1702 | return 0; | |
1703 | } else | |
1704 | option = 1; /* No argument, default to 1 */ | |
ac27a0ec | 1705 | if (option) |
fd8c37ec | 1706 | set_opt(sb, BARRIER); |
ac27a0ec | 1707 | else |
fd8c37ec | 1708 | clear_opt(sb, BARRIER); |
ac27a0ec DK |
1709 | break; |
1710 | case Opt_ignore: | |
1711 | break; | |
1712 | case Opt_resize: | |
1713 | if (!is_remount) { | |
b31e1552 ES |
1714 | ext4_msg(sb, KERN_ERR, |
1715 | "resize option only available " | |
1716 | "for remount"); | |
ac27a0ec DK |
1717 | return 0; |
1718 | } | |
1719 | if (match_int(&args[0], &option) != 0) | |
1720 | return 0; | |
1721 | *n_blocks_count = option; | |
1722 | break; | |
1723 | case Opt_nobh: | |
206f7ab4 CH |
1724 | ext4_msg(sb, KERN_WARNING, |
1725 | "Ignoring deprecated nobh option"); | |
ac27a0ec DK |
1726 | break; |
1727 | case Opt_bh: | |
206f7ab4 CH |
1728 | ext4_msg(sb, KERN_WARNING, |
1729 | "Ignoring deprecated bh option"); | |
ac27a0ec | 1730 | break; |
25ec56b5 | 1731 | case Opt_i_version: |
fd8c37ec | 1732 | set_opt(sb, I_VERSION); |
25ec56b5 JNC |
1733 | sb->s_flags |= MS_I_VERSION; |
1734 | break; | |
dd919b98 | 1735 | case Opt_nodelalloc: |
fd8c37ec | 1736 | clear_opt(sb, DELALLOC); |
dd919b98 | 1737 | break; |
1449032b | 1738 | case Opt_mblk_io_submit: |
fd8c37ec | 1739 | set_opt(sb, MBLK_IO_SUBMIT); |
1449032b TT |
1740 | break; |
1741 | case Opt_nomblk_io_submit: | |
fd8c37ec | 1742 | clear_opt(sb, MBLK_IO_SUBMIT); |
1449032b | 1743 | break; |
c9de560d AT |
1744 | case Opt_stripe: |
1745 | if (match_int(&args[0], &option)) | |
1746 | return 0; | |
1747 | if (option < 0) | |
1748 | return 0; | |
1749 | sbi->s_stripe = option; | |
1750 | break; | |
64769240 | 1751 | case Opt_delalloc: |
fd8c37ec | 1752 | set_opt(sb, DELALLOC); |
64769240 | 1753 | break; |
6fd058f7 | 1754 | case Opt_block_validity: |
fd8c37ec | 1755 | set_opt(sb, BLOCK_VALIDITY); |
6fd058f7 TT |
1756 | break; |
1757 | case Opt_noblock_validity: | |
fd8c37ec | 1758 | clear_opt(sb, BLOCK_VALIDITY); |
6fd058f7 | 1759 | break; |
240799cd TT |
1760 | case Opt_inode_readahead_blks: |
1761 | if (match_int(&args[0], &option)) | |
1762 | return 0; | |
1763 | if (option < 0 || option > (1 << 30)) | |
1764 | return 0; | |
f7c43950 | 1765 | if (!is_power_of_2(option)) { |
b31e1552 ES |
1766 | ext4_msg(sb, KERN_ERR, |
1767 | "EXT4-fs: inode_readahead_blks" | |
1768 | " must be a power of 2"); | |
3197ebdb TT |
1769 | return 0; |
1770 | } | |
240799cd TT |
1771 | sbi->s_inode_readahead_blks = option; |
1772 | break; | |
b3881f74 TT |
1773 | case Opt_journal_ioprio: |
1774 | if (match_int(&args[0], &option)) | |
1775 | return 0; | |
1776 | if (option < 0 || option > 7) | |
1777 | break; | |
1778 | *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, | |
1779 | option); | |
1780 | break; | |
06705bff | 1781 | case Opt_noauto_da_alloc: |
fd8c37ec | 1782 | set_opt(sb, NO_AUTO_DA_ALLOC); |
06705bff | 1783 | break; |
afd4672d | 1784 | case Opt_auto_da_alloc: |
15121c18 ES |
1785 | if (args[0].from) { |
1786 | if (match_int(&args[0], &option)) | |
1787 | return 0; | |
1788 | } else | |
1789 | option = 1; /* No argument, default to 1 */ | |
afd4672d | 1790 | if (option) |
fd8c37ec | 1791 | clear_opt(sb, NO_AUTO_DA_ALLOC); |
afd4672d | 1792 | else |
fd8c37ec | 1793 | set_opt(sb,NO_AUTO_DA_ALLOC); |
afd4672d | 1794 | break; |
5328e635 | 1795 | case Opt_discard: |
fd8c37ec | 1796 | set_opt(sb, DISCARD); |
5328e635 ES |
1797 | break; |
1798 | case Opt_nodiscard: | |
fd8c37ec | 1799 | clear_opt(sb, DISCARD); |
5328e635 | 1800 | break; |
744692dc | 1801 | case Opt_dioread_nolock: |
fd8c37ec | 1802 | set_opt(sb, DIOREAD_NOLOCK); |
744692dc JZ |
1803 | break; |
1804 | case Opt_dioread_lock: | |
fd8c37ec | 1805 | clear_opt(sb, DIOREAD_NOLOCK); |
744692dc | 1806 | break; |
bfff6873 | 1807 | case Opt_init_inode_table: |
fd8c37ec | 1808 | set_opt(sb, INIT_INODE_TABLE); |
bfff6873 LC |
1809 | if (args[0].from) { |
1810 | if (match_int(&args[0], &option)) | |
1811 | return 0; | |
1812 | } else | |
1813 | option = EXT4_DEF_LI_WAIT_MULT; | |
1814 | if (option < 0) | |
1815 | return 0; | |
1816 | sbi->s_li_wait_mult = option; | |
1817 | break; | |
1818 | case Opt_noinit_inode_table: | |
fd8c37ec | 1819 | clear_opt(sb, INIT_INODE_TABLE); |
bfff6873 | 1820 | break; |
ac27a0ec | 1821 | default: |
b31e1552 ES |
1822 | ext4_msg(sb, KERN_ERR, |
1823 | "Unrecognized mount option \"%s\" " | |
1824 | "or missing value", p); | |
ac27a0ec DK |
1825 | return 0; |
1826 | } | |
1827 | } | |
1828 | #ifdef CONFIG_QUOTA | |
1829 | if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { | |
482a7425 | 1830 | if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA]) |
fd8c37ec | 1831 | clear_opt(sb, USRQUOTA); |
ac27a0ec | 1832 | |
482a7425 | 1833 | if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA]) |
fd8c37ec | 1834 | clear_opt(sb, GRPQUOTA); |
ac27a0ec | 1835 | |
56c50f11 | 1836 | if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { |
b31e1552 ES |
1837 | ext4_msg(sb, KERN_ERR, "old and new quota " |
1838 | "format mixing"); | |
ac27a0ec DK |
1839 | return 0; |
1840 | } | |
1841 | ||
1842 | if (!sbi->s_jquota_fmt) { | |
b31e1552 ES |
1843 | ext4_msg(sb, KERN_ERR, "journaled quota format " |
1844 | "not specified"); | |
ac27a0ec DK |
1845 | return 0; |
1846 | } | |
1847 | } else { | |
1848 | if (sbi->s_jquota_fmt) { | |
b31e1552 | 1849 | ext4_msg(sb, KERN_ERR, "journaled quota format " |
2c8be6b2 | 1850 | "specified with no journaling " |
b31e1552 | 1851 | "enabled"); |
ac27a0ec DK |
1852 | return 0; |
1853 | } | |
1854 | } | |
1855 | #endif | |
1856 | return 1; | |
1857 | } | |
1858 | ||
617ba13b | 1859 | static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, |
ac27a0ec DK |
1860 | int read_only) |
1861 | { | |
617ba13b | 1862 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec DK |
1863 | int res = 0; |
1864 | ||
617ba13b | 1865 | if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { |
b31e1552 ES |
1866 | ext4_msg(sb, KERN_ERR, "revision level too high, " |
1867 | "forcing read-only mode"); | |
ac27a0ec DK |
1868 | res = MS_RDONLY; |
1869 | } | |
1870 | if (read_only) | |
1871 | return res; | |
617ba13b | 1872 | if (!(sbi->s_mount_state & EXT4_VALID_FS)) |
b31e1552 ES |
1873 | ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " |
1874 | "running e2fsck is recommended"); | |
617ba13b | 1875 | else if ((sbi->s_mount_state & EXT4_ERROR_FS)) |
b31e1552 ES |
1876 | ext4_msg(sb, KERN_WARNING, |
1877 | "warning: mounting fs with errors, " | |
1878 | "running e2fsck is recommended"); | |
ac27a0ec DK |
1879 | else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 && |
1880 | le16_to_cpu(es->s_mnt_count) >= | |
1881 | (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) | |
b31e1552 ES |
1882 | ext4_msg(sb, KERN_WARNING, |
1883 | "warning: maximal mount count reached, " | |
1884 | "running e2fsck is recommended"); | |
ac27a0ec DK |
1885 | else if (le32_to_cpu(es->s_checkinterval) && |
1886 | (le32_to_cpu(es->s_lastcheck) + | |
1887 | le32_to_cpu(es->s_checkinterval) <= get_seconds())) | |
b31e1552 ES |
1888 | ext4_msg(sb, KERN_WARNING, |
1889 | "warning: checktime reached, " | |
1890 | "running e2fsck is recommended"); | |
0b8e58a1 | 1891 | if (!sbi->s_journal) |
0390131b | 1892 | es->s_state &= cpu_to_le16(~EXT4_VALID_FS); |
ac27a0ec | 1893 | if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) |
617ba13b | 1894 | es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); |
e8546d06 | 1895 | le16_add_cpu(&es->s_mnt_count, 1); |
ac27a0ec | 1896 | es->s_mtime = cpu_to_le32(get_seconds()); |
617ba13b | 1897 | ext4_update_dynamic_rev(sb); |
0390131b FM |
1898 | if (sbi->s_journal) |
1899 | EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
ac27a0ec | 1900 | |
e2d67052 | 1901 | ext4_commit_super(sb, 1); |
ac27a0ec | 1902 | if (test_opt(sb, DEBUG)) |
a9df9a49 | 1903 | printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " |
a2595b8a | 1904 | "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", |
ac27a0ec DK |
1905 | sb->s_blocksize, |
1906 | sbi->s_groups_count, | |
617ba13b MC |
1907 | EXT4_BLOCKS_PER_GROUP(sb), |
1908 | EXT4_INODES_PER_GROUP(sb), | |
a2595b8a | 1909 | sbi->s_mount_opt, sbi->s_mount_opt2); |
ac27a0ec | 1910 | |
ac27a0ec DK |
1911 | return res; |
1912 | } | |
1913 | ||
772cb7c8 JS |
1914 | static int ext4_fill_flex_info(struct super_block *sb) |
1915 | { | |
1916 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1917 | struct ext4_group_desc *gdp = NULL; | |
772cb7c8 JS |
1918 | ext4_group_t flex_group_count; |
1919 | ext4_group_t flex_group; | |
1920 | int groups_per_flex = 0; | |
c5ca7c76 | 1921 | size_t size; |
772cb7c8 JS |
1922 | int i; |
1923 | ||
503358ae TT |
1924 | sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; |
1925 | groups_per_flex = 1 << sbi->s_log_groups_per_flex; | |
1926 | ||
1927 | if (groups_per_flex < 2) { | |
772cb7c8 JS |
1928 | sbi->s_log_groups_per_flex = 0; |
1929 | return 1; | |
1930 | } | |
1931 | ||
c62a11fd FB |
1932 | /* We allocate both existing and potentially added groups */ |
1933 | flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) + | |
d94e99a6 AK |
1934 | ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) << |
1935 | EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex; | |
c5ca7c76 TT |
1936 | size = flex_group_count * sizeof(struct flex_groups); |
1937 | sbi->s_flex_groups = kzalloc(size, GFP_KERNEL); | |
1938 | if (sbi->s_flex_groups == NULL) { | |
94de56ab JP |
1939 | sbi->s_flex_groups = vzalloc(size); |
1940 | if (sbi->s_flex_groups == NULL) { | |
1941 | ext4_msg(sb, KERN_ERR, | |
1942 | "not enough memory for %u flex groups", | |
1943 | flex_group_count); | |
1944 | goto failed; | |
1945 | } | |
772cb7c8 | 1946 | } |
772cb7c8 | 1947 | |
772cb7c8 | 1948 | for (i = 0; i < sbi->s_groups_count; i++) { |
88b6edd1 | 1949 | gdp = ext4_get_group_desc(sb, i, NULL); |
772cb7c8 JS |
1950 | |
1951 | flex_group = ext4_flex_group(sbi, i); | |
7ad9bb65 TT |
1952 | atomic_add(ext4_free_inodes_count(sb, gdp), |
1953 | &sbi->s_flex_groups[flex_group].free_inodes); | |
1954 | atomic_add(ext4_free_blks_count(sb, gdp), | |
1955 | &sbi->s_flex_groups[flex_group].free_blocks); | |
1956 | atomic_add(ext4_used_dirs_count(sb, gdp), | |
1957 | &sbi->s_flex_groups[flex_group].used_dirs); | |
772cb7c8 JS |
1958 | } |
1959 | ||
1960 | return 1; | |
1961 | failed: | |
1962 | return 0; | |
1963 | } | |
1964 | ||
717d50e4 AD |
1965 | __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group, |
1966 | struct ext4_group_desc *gdp) | |
1967 | { | |
1968 | __u16 crc = 0; | |
1969 | ||
1970 | if (sbi->s_es->s_feature_ro_compat & | |
1971 | cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) { | |
1972 | int offset = offsetof(struct ext4_group_desc, bg_checksum); | |
1973 | __le32 le_group = cpu_to_le32(block_group); | |
1974 | ||
1975 | crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); | |
1976 | crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); | |
1977 | crc = crc16(crc, (__u8 *)gdp, offset); | |
1978 | offset += sizeof(gdp->bg_checksum); /* skip checksum */ | |
1979 | /* for checksum of struct ext4_group_desc do the rest...*/ | |
1980 | if ((sbi->s_es->s_feature_incompat & | |
1981 | cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) && | |
1982 | offset < le16_to_cpu(sbi->s_es->s_desc_size)) | |
1983 | crc = crc16(crc, (__u8 *)gdp + offset, | |
1984 | le16_to_cpu(sbi->s_es->s_desc_size) - | |
1985 | offset); | |
1986 | } | |
1987 | ||
1988 | return cpu_to_le16(crc); | |
1989 | } | |
1990 | ||
1991 | int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group, | |
1992 | struct ext4_group_desc *gdp) | |
1993 | { | |
1994 | if ((sbi->s_es->s_feature_ro_compat & | |
1995 | cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) && | |
1996 | (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp))) | |
1997 | return 0; | |
1998 | ||
1999 | return 1; | |
2000 | } | |
2001 | ||
ac27a0ec | 2002 | /* Called at mount-time, super-block is locked */ |
bfff6873 LC |
2003 | static int ext4_check_descriptors(struct super_block *sb, |
2004 | ext4_group_t *first_not_zeroed) | |
ac27a0ec | 2005 | { |
617ba13b MC |
2006 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
2007 | ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); | |
2008 | ext4_fsblk_t last_block; | |
bd81d8ee LV |
2009 | ext4_fsblk_t block_bitmap; |
2010 | ext4_fsblk_t inode_bitmap; | |
2011 | ext4_fsblk_t inode_table; | |
ce421581 | 2012 | int flexbg_flag = 0; |
bfff6873 | 2013 | ext4_group_t i, grp = sbi->s_groups_count; |
ac27a0ec | 2014 | |
ce421581 JS |
2015 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) |
2016 | flexbg_flag = 1; | |
2017 | ||
af5bc92d | 2018 | ext4_debug("Checking group descriptors"); |
ac27a0ec | 2019 | |
197cd65a AM |
2020 | for (i = 0; i < sbi->s_groups_count; i++) { |
2021 | struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); | |
2022 | ||
ce421581 | 2023 | if (i == sbi->s_groups_count - 1 || flexbg_flag) |
bd81d8ee | 2024 | last_block = ext4_blocks_count(sbi->s_es) - 1; |
ac27a0ec DK |
2025 | else |
2026 | last_block = first_block + | |
617ba13b | 2027 | (EXT4_BLOCKS_PER_GROUP(sb) - 1); |
ac27a0ec | 2028 | |
bfff6873 LC |
2029 | if ((grp == sbi->s_groups_count) && |
2030 | !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) | |
2031 | grp = i; | |
2032 | ||
8fadc143 | 2033 | block_bitmap = ext4_block_bitmap(sb, gdp); |
2b2d6d01 | 2034 | if (block_bitmap < first_block || block_bitmap > last_block) { |
b31e1552 | 2035 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 2036 | "Block bitmap for group %u not in group " |
b31e1552 | 2037 | "(block %llu)!", i, block_bitmap); |
ac27a0ec DK |
2038 | return 0; |
2039 | } | |
8fadc143 | 2040 | inode_bitmap = ext4_inode_bitmap(sb, gdp); |
2b2d6d01 | 2041 | if (inode_bitmap < first_block || inode_bitmap > last_block) { |
b31e1552 | 2042 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 2043 | "Inode bitmap for group %u not in group " |
b31e1552 | 2044 | "(block %llu)!", i, inode_bitmap); |
ac27a0ec DK |
2045 | return 0; |
2046 | } | |
8fadc143 | 2047 | inode_table = ext4_inode_table(sb, gdp); |
bd81d8ee | 2048 | if (inode_table < first_block || |
2b2d6d01 | 2049 | inode_table + sbi->s_itb_per_group - 1 > last_block) { |
b31e1552 | 2050 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 2051 | "Inode table for group %u not in group " |
b31e1552 | 2052 | "(block %llu)!", i, inode_table); |
ac27a0ec DK |
2053 | return 0; |
2054 | } | |
955ce5f5 | 2055 | ext4_lock_group(sb, i); |
717d50e4 | 2056 | if (!ext4_group_desc_csum_verify(sbi, i, gdp)) { |
b31e1552 ES |
2057 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
2058 | "Checksum for group %u failed (%u!=%u)", | |
2059 | i, le16_to_cpu(ext4_group_desc_csum(sbi, i, | |
2060 | gdp)), le16_to_cpu(gdp->bg_checksum)); | |
7ee1ec4c | 2061 | if (!(sb->s_flags & MS_RDONLY)) { |
955ce5f5 | 2062 | ext4_unlock_group(sb, i); |
8a266467 | 2063 | return 0; |
7ee1ec4c | 2064 | } |
717d50e4 | 2065 | } |
955ce5f5 | 2066 | ext4_unlock_group(sb, i); |
ce421581 JS |
2067 | if (!flexbg_flag) |
2068 | first_block += EXT4_BLOCKS_PER_GROUP(sb); | |
ac27a0ec | 2069 | } |
bfff6873 LC |
2070 | if (NULL != first_not_zeroed) |
2071 | *first_not_zeroed = grp; | |
ac27a0ec | 2072 | |
bd81d8ee | 2073 | ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb)); |
0b8e58a1 | 2074 | sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb)); |
ac27a0ec DK |
2075 | return 1; |
2076 | } | |
2077 | ||
617ba13b | 2078 | /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at |
ac27a0ec DK |
2079 | * the superblock) which were deleted from all directories, but held open by |
2080 | * a process at the time of a crash. We walk the list and try to delete these | |
2081 | * inodes at recovery time (only with a read-write filesystem). | |
2082 | * | |
2083 | * In order to keep the orphan inode chain consistent during traversal (in | |
2084 | * case of crash during recovery), we link each inode into the superblock | |
2085 | * orphan list_head and handle it the same way as an inode deletion during | |
2086 | * normal operation (which journals the operations for us). | |
2087 | * | |
2088 | * We only do an iget() and an iput() on each inode, which is very safe if we | |
2089 | * accidentally point at an in-use or already deleted inode. The worst that | |
2090 | * can happen in this case is that we get a "bit already cleared" message from | |
617ba13b | 2091 | * ext4_free_inode(). The only reason we would point at a wrong inode is if |
ac27a0ec DK |
2092 | * e2fsck was run on this filesystem, and it must have already done the orphan |
2093 | * inode cleanup for us, so we can safely abort without any further action. | |
2094 | */ | |
2b2d6d01 TT |
2095 | static void ext4_orphan_cleanup(struct super_block *sb, |
2096 | struct ext4_super_block *es) | |
ac27a0ec DK |
2097 | { |
2098 | unsigned int s_flags = sb->s_flags; | |
2099 | int nr_orphans = 0, nr_truncates = 0; | |
2100 | #ifdef CONFIG_QUOTA | |
2101 | int i; | |
2102 | #endif | |
2103 | if (!es->s_last_orphan) { | |
2104 | jbd_debug(4, "no orphan inodes to clean up\n"); | |
2105 | return; | |
2106 | } | |
2107 | ||
a8f48a95 | 2108 | if (bdev_read_only(sb->s_bdev)) { |
b31e1552 ES |
2109 | ext4_msg(sb, KERN_ERR, "write access " |
2110 | "unavailable, skipping orphan cleanup"); | |
a8f48a95 ES |
2111 | return; |
2112 | } | |
2113 | ||
617ba13b | 2114 | if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { |
ac27a0ec DK |
2115 | if (es->s_last_orphan) |
2116 | jbd_debug(1, "Errors on filesystem, " | |
2117 | "clearing orphan list.\n"); | |
2118 | es->s_last_orphan = 0; | |
2119 | jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); | |
2120 | return; | |
2121 | } | |
2122 | ||
2123 | if (s_flags & MS_RDONLY) { | |
b31e1552 | 2124 | ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); |
ac27a0ec DK |
2125 | sb->s_flags &= ~MS_RDONLY; |
2126 | } | |
2127 | #ifdef CONFIG_QUOTA | |
2128 | /* Needed for iput() to work correctly and not trash data */ | |
2129 | sb->s_flags |= MS_ACTIVE; | |
2130 | /* Turn on quotas so that they are updated correctly */ | |
2131 | for (i = 0; i < MAXQUOTAS; i++) { | |
617ba13b MC |
2132 | if (EXT4_SB(sb)->s_qf_names[i]) { |
2133 | int ret = ext4_quota_on_mount(sb, i); | |
ac27a0ec | 2134 | if (ret < 0) |
b31e1552 ES |
2135 | ext4_msg(sb, KERN_ERR, |
2136 | "Cannot turn on journaled " | |
2137 | "quota: error %d", ret); | |
ac27a0ec DK |
2138 | } |
2139 | } | |
2140 | #endif | |
2141 | ||
2142 | while (es->s_last_orphan) { | |
2143 | struct inode *inode; | |
2144 | ||
97bd42b9 JB |
2145 | inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); |
2146 | if (IS_ERR(inode)) { | |
ac27a0ec DK |
2147 | es->s_last_orphan = 0; |
2148 | break; | |
2149 | } | |
2150 | ||
617ba13b | 2151 | list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); |
871a2931 | 2152 | dquot_initialize(inode); |
ac27a0ec | 2153 | if (inode->i_nlink) { |
b31e1552 ES |
2154 | ext4_msg(sb, KERN_DEBUG, |
2155 | "%s: truncating inode %lu to %lld bytes", | |
46e665e9 | 2156 | __func__, inode->i_ino, inode->i_size); |
e5f8eab8 | 2157 | jbd_debug(2, "truncating inode %lu to %lld bytes\n", |
ac27a0ec | 2158 | inode->i_ino, inode->i_size); |
617ba13b | 2159 | ext4_truncate(inode); |
ac27a0ec DK |
2160 | nr_truncates++; |
2161 | } else { | |
b31e1552 ES |
2162 | ext4_msg(sb, KERN_DEBUG, |
2163 | "%s: deleting unreferenced inode %lu", | |
46e665e9 | 2164 | __func__, inode->i_ino); |
ac27a0ec DK |
2165 | jbd_debug(2, "deleting unreferenced inode %lu\n", |
2166 | inode->i_ino); | |
2167 | nr_orphans++; | |
2168 | } | |
2169 | iput(inode); /* The delete magic happens here! */ | |
2170 | } | |
2171 | ||
2b2d6d01 | 2172 | #define PLURAL(x) (x), ((x) == 1) ? "" : "s" |
ac27a0ec DK |
2173 | |
2174 | if (nr_orphans) | |
b31e1552 ES |
2175 | ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", |
2176 | PLURAL(nr_orphans)); | |
ac27a0ec | 2177 | if (nr_truncates) |
b31e1552 ES |
2178 | ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", |
2179 | PLURAL(nr_truncates)); | |
ac27a0ec DK |
2180 | #ifdef CONFIG_QUOTA |
2181 | /* Turn quotas off */ | |
2182 | for (i = 0; i < MAXQUOTAS; i++) { | |
2183 | if (sb_dqopt(sb)->files[i]) | |
287a8095 | 2184 | dquot_quota_off(sb, i); |
ac27a0ec DK |
2185 | } |
2186 | #endif | |
2187 | sb->s_flags = s_flags; /* Restore MS_RDONLY status */ | |
2188 | } | |
0b8e58a1 | 2189 | |
cd2291a4 ES |
2190 | /* |
2191 | * Maximal extent format file size. | |
2192 | * Resulting logical blkno at s_maxbytes must fit in our on-disk | |
2193 | * extent format containers, within a sector_t, and within i_blocks | |
2194 | * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, | |
2195 | * so that won't be a limiting factor. | |
2196 | * | |
2197 | * Note, this does *not* consider any metadata overhead for vfs i_blocks. | |
2198 | */ | |
f287a1a5 | 2199 | static loff_t ext4_max_size(int blkbits, int has_huge_files) |
cd2291a4 ES |
2200 | { |
2201 | loff_t res; | |
2202 | loff_t upper_limit = MAX_LFS_FILESIZE; | |
2203 | ||
2204 | /* small i_blocks in vfs inode? */ | |
f287a1a5 | 2205 | if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { |
cd2291a4 | 2206 | /* |
90c699a9 | 2207 | * CONFIG_LBDAF is not enabled implies the inode |
cd2291a4 ES |
2208 | * i_block represent total blocks in 512 bytes |
2209 | * 32 == size of vfs inode i_blocks * 8 | |
2210 | */ | |
2211 | upper_limit = (1LL << 32) - 1; | |
2212 | ||
2213 | /* total blocks in file system block size */ | |
2214 | upper_limit >>= (blkbits - 9); | |
2215 | upper_limit <<= blkbits; | |
2216 | } | |
2217 | ||
2218 | /* 32-bit extent-start container, ee_block */ | |
2219 | res = 1LL << 32; | |
2220 | res <<= blkbits; | |
2221 | res -= 1; | |
2222 | ||
2223 | /* Sanity check against vm- & vfs- imposed limits */ | |
2224 | if (res > upper_limit) | |
2225 | res = upper_limit; | |
2226 | ||
2227 | return res; | |
2228 | } | |
ac27a0ec | 2229 | |
ac27a0ec | 2230 | /* |
cd2291a4 | 2231 | * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect |
0fc1b451 AK |
2232 | * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. |
2233 | * We need to be 1 filesystem block less than the 2^48 sector limit. | |
ac27a0ec | 2234 | */ |
f287a1a5 | 2235 | static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) |
ac27a0ec | 2236 | { |
617ba13b | 2237 | loff_t res = EXT4_NDIR_BLOCKS; |
0fc1b451 AK |
2238 | int meta_blocks; |
2239 | loff_t upper_limit; | |
0b8e58a1 AD |
2240 | /* This is calculated to be the largest file size for a dense, block |
2241 | * mapped file such that the file's total number of 512-byte sectors, | |
2242 | * including data and all indirect blocks, does not exceed (2^48 - 1). | |
2243 | * | |
2244 | * __u32 i_blocks_lo and _u16 i_blocks_high represent the total | |
2245 | * number of 512-byte sectors of the file. | |
0fc1b451 AK |
2246 | */ |
2247 | ||
f287a1a5 | 2248 | if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { |
0fc1b451 | 2249 | /* |
90c699a9 | 2250 | * !has_huge_files or CONFIG_LBDAF not enabled implies that |
0b8e58a1 AD |
2251 | * the inode i_block field represents total file blocks in |
2252 | * 2^32 512-byte sectors == size of vfs inode i_blocks * 8 | |
0fc1b451 AK |
2253 | */ |
2254 | upper_limit = (1LL << 32) - 1; | |
2255 | ||
2256 | /* total blocks in file system block size */ | |
2257 | upper_limit >>= (bits - 9); | |
2258 | ||
2259 | } else { | |
8180a562 AK |
2260 | /* |
2261 | * We use 48 bit ext4_inode i_blocks | |
2262 | * With EXT4_HUGE_FILE_FL set the i_blocks | |
2263 | * represent total number of blocks in | |
2264 | * file system block size | |
2265 | */ | |
0fc1b451 AK |
2266 | upper_limit = (1LL << 48) - 1; |
2267 | ||
0fc1b451 AK |
2268 | } |
2269 | ||
2270 | /* indirect blocks */ | |
2271 | meta_blocks = 1; | |
2272 | /* double indirect blocks */ | |
2273 | meta_blocks += 1 + (1LL << (bits-2)); | |
2274 | /* tripple indirect blocks */ | |
2275 | meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); | |
2276 | ||
2277 | upper_limit -= meta_blocks; | |
2278 | upper_limit <<= bits; | |
ac27a0ec DK |
2279 | |
2280 | res += 1LL << (bits-2); | |
2281 | res += 1LL << (2*(bits-2)); | |
2282 | res += 1LL << (3*(bits-2)); | |
2283 | res <<= bits; | |
2284 | if (res > upper_limit) | |
2285 | res = upper_limit; | |
0fc1b451 AK |
2286 | |
2287 | if (res > MAX_LFS_FILESIZE) | |
2288 | res = MAX_LFS_FILESIZE; | |
2289 | ||
ac27a0ec DK |
2290 | return res; |
2291 | } | |
2292 | ||
617ba13b | 2293 | static ext4_fsblk_t descriptor_loc(struct super_block *sb, |
0b8e58a1 | 2294 | ext4_fsblk_t logical_sb_block, int nr) |
ac27a0ec | 2295 | { |
617ba13b | 2296 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
fd2d4291 | 2297 | ext4_group_t bg, first_meta_bg; |
ac27a0ec DK |
2298 | int has_super = 0; |
2299 | ||
2300 | first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); | |
2301 | ||
617ba13b | 2302 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) || |
ac27a0ec | 2303 | nr < first_meta_bg) |
70bbb3e0 | 2304 | return logical_sb_block + nr + 1; |
ac27a0ec | 2305 | bg = sbi->s_desc_per_block * nr; |
617ba13b | 2306 | if (ext4_bg_has_super(sb, bg)) |
ac27a0ec | 2307 | has_super = 1; |
0b8e58a1 | 2308 | |
617ba13b | 2309 | return (has_super + ext4_group_first_block_no(sb, bg)); |
ac27a0ec DK |
2310 | } |
2311 | ||
c9de560d AT |
2312 | /** |
2313 | * ext4_get_stripe_size: Get the stripe size. | |
2314 | * @sbi: In memory super block info | |
2315 | * | |
2316 | * If we have specified it via mount option, then | |
2317 | * use the mount option value. If the value specified at mount time is | |
2318 | * greater than the blocks per group use the super block value. | |
2319 | * If the super block value is greater than blocks per group return 0. | |
2320 | * Allocator needs it be less than blocks per group. | |
2321 | * | |
2322 | */ | |
2323 | static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) | |
2324 | { | |
2325 | unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); | |
2326 | unsigned long stripe_width = | |
2327 | le32_to_cpu(sbi->s_es->s_raid_stripe_width); | |
2328 | ||
2329 | if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) | |
2330 | return sbi->s_stripe; | |
2331 | ||
2332 | if (stripe_width <= sbi->s_blocks_per_group) | |
2333 | return stripe_width; | |
2334 | ||
2335 | if (stride <= sbi->s_blocks_per_group) | |
2336 | return stride; | |
2337 | ||
2338 | return 0; | |
2339 | } | |
ac27a0ec | 2340 | |
3197ebdb TT |
2341 | /* sysfs supprt */ |
2342 | ||
2343 | struct ext4_attr { | |
2344 | struct attribute attr; | |
2345 | ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *); | |
60e6679e | 2346 | ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *, |
3197ebdb TT |
2347 | const char *, size_t); |
2348 | int offset; | |
2349 | }; | |
2350 | ||
2351 | static int parse_strtoul(const char *buf, | |
2352 | unsigned long max, unsigned long *value) | |
2353 | { | |
2354 | char *endp; | |
2355 | ||
e7d2860b AGR |
2356 | *value = simple_strtoul(skip_spaces(buf), &endp, 0); |
2357 | endp = skip_spaces(endp); | |
3197ebdb TT |
2358 | if (*endp || *value > max) |
2359 | return -EINVAL; | |
2360 | ||
2361 | return 0; | |
2362 | } | |
2363 | ||
2364 | static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a, | |
2365 | struct ext4_sb_info *sbi, | |
2366 | char *buf) | |
2367 | { | |
2368 | return snprintf(buf, PAGE_SIZE, "%llu\n", | |
2369 | (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter)); | |
2370 | } | |
2371 | ||
2372 | static ssize_t session_write_kbytes_show(struct ext4_attr *a, | |
2373 | struct ext4_sb_info *sbi, char *buf) | |
2374 | { | |
2375 | struct super_block *sb = sbi->s_buddy_cache->i_sb; | |
2376 | ||
f613dfcb TT |
2377 | if (!sb->s_bdev->bd_part) |
2378 | return snprintf(buf, PAGE_SIZE, "0\n"); | |
3197ebdb TT |
2379 | return snprintf(buf, PAGE_SIZE, "%lu\n", |
2380 | (part_stat_read(sb->s_bdev->bd_part, sectors[1]) - | |
2381 | sbi->s_sectors_written_start) >> 1); | |
2382 | } | |
2383 | ||
2384 | static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a, | |
2385 | struct ext4_sb_info *sbi, char *buf) | |
2386 | { | |
2387 | struct super_block *sb = sbi->s_buddy_cache->i_sb; | |
2388 | ||
f613dfcb TT |
2389 | if (!sb->s_bdev->bd_part) |
2390 | return snprintf(buf, PAGE_SIZE, "0\n"); | |
3197ebdb | 2391 | return snprintf(buf, PAGE_SIZE, "%llu\n", |
a6b43e38 | 2392 | (unsigned long long)(sbi->s_kbytes_written + |
3197ebdb | 2393 | ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - |
a6b43e38 | 2394 | EXT4_SB(sb)->s_sectors_written_start) >> 1))); |
3197ebdb TT |
2395 | } |
2396 | ||
2397 | static ssize_t inode_readahead_blks_store(struct ext4_attr *a, | |
2398 | struct ext4_sb_info *sbi, | |
2399 | const char *buf, size_t count) | |
2400 | { | |
2401 | unsigned long t; | |
2402 | ||
2403 | if (parse_strtoul(buf, 0x40000000, &t)) | |
2404 | return -EINVAL; | |
2405 | ||
f7c43950 | 2406 | if (!is_power_of_2(t)) |
3197ebdb TT |
2407 | return -EINVAL; |
2408 | ||
2409 | sbi->s_inode_readahead_blks = t; | |
2410 | return count; | |
2411 | } | |
2412 | ||
2413 | static ssize_t sbi_ui_show(struct ext4_attr *a, | |
0b8e58a1 | 2414 | struct ext4_sb_info *sbi, char *buf) |
3197ebdb TT |
2415 | { |
2416 | unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset); | |
2417 | ||
2418 | return snprintf(buf, PAGE_SIZE, "%u\n", *ui); | |
2419 | } | |
2420 | ||
2421 | static ssize_t sbi_ui_store(struct ext4_attr *a, | |
2422 | struct ext4_sb_info *sbi, | |
2423 | const char *buf, size_t count) | |
2424 | { | |
2425 | unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset); | |
2426 | unsigned long t; | |
2427 | ||
2428 | if (parse_strtoul(buf, 0xffffffff, &t)) | |
2429 | return -EINVAL; | |
2430 | *ui = t; | |
2431 | return count; | |
2432 | } | |
2433 | ||
2434 | #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \ | |
2435 | static struct ext4_attr ext4_attr_##_name = { \ | |
2436 | .attr = {.name = __stringify(_name), .mode = _mode }, \ | |
2437 | .show = _show, \ | |
2438 | .store = _store, \ | |
2439 | .offset = offsetof(struct ext4_sb_info, _elname), \ | |
2440 | } | |
2441 | #define EXT4_ATTR(name, mode, show, store) \ | |
2442 | static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store) | |
2443 | ||
857ac889 | 2444 | #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL) |
3197ebdb TT |
2445 | #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL) |
2446 | #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store) | |
2447 | #define EXT4_RW_ATTR_SBI_UI(name, elname) \ | |
2448 | EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname) | |
2449 | #define ATTR_LIST(name) &ext4_attr_##name.attr | |
2450 | ||
2451 | EXT4_RO_ATTR(delayed_allocation_blocks); | |
2452 | EXT4_RO_ATTR(session_write_kbytes); | |
2453 | EXT4_RO_ATTR(lifetime_write_kbytes); | |
2454 | EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show, | |
2455 | inode_readahead_blks_store, s_inode_readahead_blks); | |
11013911 | 2456 | EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal); |
3197ebdb TT |
2457 | EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats); |
2458 | EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan); | |
2459 | EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan); | |
2460 | EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs); | |
2461 | EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request); | |
2462 | EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc); | |
55138e0b | 2463 | EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump); |
3197ebdb TT |
2464 | |
2465 | static struct attribute *ext4_attrs[] = { | |
2466 | ATTR_LIST(delayed_allocation_blocks), | |
2467 | ATTR_LIST(session_write_kbytes), | |
2468 | ATTR_LIST(lifetime_write_kbytes), | |
2469 | ATTR_LIST(inode_readahead_blks), | |
11013911 | 2470 | ATTR_LIST(inode_goal), |
3197ebdb TT |
2471 | ATTR_LIST(mb_stats), |
2472 | ATTR_LIST(mb_max_to_scan), | |
2473 | ATTR_LIST(mb_min_to_scan), | |
2474 | ATTR_LIST(mb_order2_req), | |
2475 | ATTR_LIST(mb_stream_req), | |
2476 | ATTR_LIST(mb_group_prealloc), | |
55138e0b | 2477 | ATTR_LIST(max_writeback_mb_bump), |
3197ebdb TT |
2478 | NULL, |
2479 | }; | |
2480 | ||
857ac889 LC |
2481 | /* Features this copy of ext4 supports */ |
2482 | EXT4_INFO_ATTR(lazy_itable_init); | |
27ee40df | 2483 | EXT4_INFO_ATTR(batched_discard); |
857ac889 LC |
2484 | |
2485 | static struct attribute *ext4_feat_attrs[] = { | |
2486 | ATTR_LIST(lazy_itable_init), | |
27ee40df | 2487 | ATTR_LIST(batched_discard), |
857ac889 LC |
2488 | NULL, |
2489 | }; | |
2490 | ||
3197ebdb TT |
2491 | static ssize_t ext4_attr_show(struct kobject *kobj, |
2492 | struct attribute *attr, char *buf) | |
2493 | { | |
2494 | struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, | |
2495 | s_kobj); | |
2496 | struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); | |
2497 | ||
2498 | return a->show ? a->show(a, sbi, buf) : 0; | |
2499 | } | |
2500 | ||
2501 | static ssize_t ext4_attr_store(struct kobject *kobj, | |
2502 | struct attribute *attr, | |
2503 | const char *buf, size_t len) | |
2504 | { | |
2505 | struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, | |
2506 | s_kobj); | |
2507 | struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); | |
2508 | ||
2509 | return a->store ? a->store(a, sbi, buf, len) : 0; | |
2510 | } | |
2511 | ||
2512 | static void ext4_sb_release(struct kobject *kobj) | |
2513 | { | |
2514 | struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, | |
2515 | s_kobj); | |
2516 | complete(&sbi->s_kobj_unregister); | |
2517 | } | |
2518 | ||
52cf25d0 | 2519 | static const struct sysfs_ops ext4_attr_ops = { |
3197ebdb TT |
2520 | .show = ext4_attr_show, |
2521 | .store = ext4_attr_store, | |
2522 | }; | |
2523 | ||
2524 | static struct kobj_type ext4_ktype = { | |
2525 | .default_attrs = ext4_attrs, | |
2526 | .sysfs_ops = &ext4_attr_ops, | |
2527 | .release = ext4_sb_release, | |
2528 | }; | |
2529 | ||
857ac889 LC |
2530 | static void ext4_feat_release(struct kobject *kobj) |
2531 | { | |
2532 | complete(&ext4_feat->f_kobj_unregister); | |
2533 | } | |
2534 | ||
2535 | static struct kobj_type ext4_feat_ktype = { | |
2536 | .default_attrs = ext4_feat_attrs, | |
2537 | .sysfs_ops = &ext4_attr_ops, | |
2538 | .release = ext4_feat_release, | |
2539 | }; | |
2540 | ||
a13fb1a4 ES |
2541 | /* |
2542 | * Check whether this filesystem can be mounted based on | |
2543 | * the features present and the RDONLY/RDWR mount requested. | |
2544 | * Returns 1 if this filesystem can be mounted as requested, | |
2545 | * 0 if it cannot be. | |
2546 | */ | |
2547 | static int ext4_feature_set_ok(struct super_block *sb, int readonly) | |
2548 | { | |
2549 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) { | |
2550 | ext4_msg(sb, KERN_ERR, | |
2551 | "Couldn't mount because of " | |
2552 | "unsupported optional features (%x)", | |
2553 | (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & | |
2554 | ~EXT4_FEATURE_INCOMPAT_SUPP)); | |
2555 | return 0; | |
2556 | } | |
2557 | ||
2558 | if (readonly) | |
2559 | return 1; | |
2560 | ||
2561 | /* Check that feature set is OK for a read-write mount */ | |
2562 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) { | |
2563 | ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " | |
2564 | "unsupported optional features (%x)", | |
2565 | (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & | |
2566 | ~EXT4_FEATURE_RO_COMPAT_SUPP)); | |
2567 | return 0; | |
2568 | } | |
2569 | /* | |
2570 | * Large file size enabled file system can only be mounted | |
2571 | * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF | |
2572 | */ | |
2573 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) { | |
2574 | if (sizeof(blkcnt_t) < sizeof(u64)) { | |
2575 | ext4_msg(sb, KERN_ERR, "Filesystem with huge files " | |
2576 | "cannot be mounted RDWR without " | |
2577 | "CONFIG_LBDAF"); | |
2578 | return 0; | |
2579 | } | |
2580 | } | |
2581 | return 1; | |
2582 | } | |
2583 | ||
66e61a9e TT |
2584 | /* |
2585 | * This function is called once a day if we have errors logged | |
2586 | * on the file system | |
2587 | */ | |
2588 | static void print_daily_error_info(unsigned long arg) | |
2589 | { | |
2590 | struct super_block *sb = (struct super_block *) arg; | |
2591 | struct ext4_sb_info *sbi; | |
2592 | struct ext4_super_block *es; | |
2593 | ||
2594 | sbi = EXT4_SB(sb); | |
2595 | es = sbi->s_es; | |
2596 | ||
2597 | if (es->s_error_count) | |
2598 | ext4_msg(sb, KERN_NOTICE, "error count: %u", | |
2599 | le32_to_cpu(es->s_error_count)); | |
2600 | if (es->s_first_error_time) { | |
2601 | printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d", | |
2602 | sb->s_id, le32_to_cpu(es->s_first_error_time), | |
2603 | (int) sizeof(es->s_first_error_func), | |
2604 | es->s_first_error_func, | |
2605 | le32_to_cpu(es->s_first_error_line)); | |
2606 | if (es->s_first_error_ino) | |
2607 | printk(": inode %u", | |
2608 | le32_to_cpu(es->s_first_error_ino)); | |
2609 | if (es->s_first_error_block) | |
2610 | printk(": block %llu", (unsigned long long) | |
2611 | le64_to_cpu(es->s_first_error_block)); | |
2612 | printk("\n"); | |
2613 | } | |
2614 | if (es->s_last_error_time) { | |
2615 | printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d", | |
2616 | sb->s_id, le32_to_cpu(es->s_last_error_time), | |
2617 | (int) sizeof(es->s_last_error_func), | |
2618 | es->s_last_error_func, | |
2619 | le32_to_cpu(es->s_last_error_line)); | |
2620 | if (es->s_last_error_ino) | |
2621 | printk(": inode %u", | |
2622 | le32_to_cpu(es->s_last_error_ino)); | |
2623 | if (es->s_last_error_block) | |
2624 | printk(": block %llu", (unsigned long long) | |
2625 | le64_to_cpu(es->s_last_error_block)); | |
2626 | printk("\n"); | |
2627 | } | |
2628 | mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ | |
2629 | } | |
2630 | ||
bfff6873 LC |
2631 | static void ext4_lazyinode_timeout(unsigned long data) |
2632 | { | |
2633 | struct task_struct *p = (struct task_struct *)data; | |
2634 | wake_up_process(p); | |
2635 | } | |
2636 | ||
2637 | /* Find next suitable group and run ext4_init_inode_table */ | |
2638 | static int ext4_run_li_request(struct ext4_li_request *elr) | |
2639 | { | |
2640 | struct ext4_group_desc *gdp = NULL; | |
2641 | ext4_group_t group, ngroups; | |
2642 | struct super_block *sb; | |
2643 | unsigned long timeout = 0; | |
2644 | int ret = 0; | |
2645 | ||
2646 | sb = elr->lr_super; | |
2647 | ngroups = EXT4_SB(sb)->s_groups_count; | |
2648 | ||
2649 | for (group = elr->lr_next_group; group < ngroups; group++) { | |
2650 | gdp = ext4_get_group_desc(sb, group, NULL); | |
2651 | if (!gdp) { | |
2652 | ret = 1; | |
2653 | break; | |
2654 | } | |
2655 | ||
2656 | if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) | |
2657 | break; | |
2658 | } | |
2659 | ||
2660 | if (group == ngroups) | |
2661 | ret = 1; | |
2662 | ||
2663 | if (!ret) { | |
2664 | timeout = jiffies; | |
2665 | ret = ext4_init_inode_table(sb, group, | |
2666 | elr->lr_timeout ? 0 : 1); | |
2667 | if (elr->lr_timeout == 0) { | |
2668 | timeout = jiffies - timeout; | |
2669 | if (elr->lr_sbi->s_li_wait_mult) | |
2670 | timeout *= elr->lr_sbi->s_li_wait_mult; | |
2671 | else | |
2672 | timeout *= 20; | |
2673 | elr->lr_timeout = timeout; | |
2674 | } | |
2675 | elr->lr_next_sched = jiffies + elr->lr_timeout; | |
2676 | elr->lr_next_group = group + 1; | |
2677 | } | |
2678 | ||
2679 | return ret; | |
2680 | } | |
2681 | ||
2682 | /* | |
2683 | * Remove lr_request from the list_request and free the | |
2684 | * request tructure. Should be called with li_list_mtx held | |
2685 | */ | |
2686 | static void ext4_remove_li_request(struct ext4_li_request *elr) | |
2687 | { | |
2688 | struct ext4_sb_info *sbi; | |
2689 | ||
2690 | if (!elr) | |
2691 | return; | |
2692 | ||
2693 | sbi = elr->lr_sbi; | |
2694 | ||
2695 | list_del(&elr->lr_request); | |
2696 | sbi->s_li_request = NULL; | |
2697 | kfree(elr); | |
2698 | } | |
2699 | ||
2700 | static void ext4_unregister_li_request(struct super_block *sb) | |
2701 | { | |
2702 | struct ext4_li_request *elr = EXT4_SB(sb)->s_li_request; | |
2703 | ||
2704 | if (!ext4_li_info) | |
2705 | return; | |
2706 | ||
2707 | mutex_lock(&ext4_li_info->li_list_mtx); | |
2708 | ext4_remove_li_request(elr); | |
2709 | mutex_unlock(&ext4_li_info->li_list_mtx); | |
2710 | } | |
2711 | ||
2712 | /* | |
2713 | * This is the function where ext4lazyinit thread lives. It walks | |
2714 | * through the request list searching for next scheduled filesystem. | |
2715 | * When such a fs is found, run the lazy initialization request | |
2716 | * (ext4_rn_li_request) and keep track of the time spend in this | |
2717 | * function. Based on that time we compute next schedule time of | |
2718 | * the request. When walking through the list is complete, compute | |
2719 | * next waking time and put itself into sleep. | |
2720 | */ | |
2721 | static int ext4_lazyinit_thread(void *arg) | |
2722 | { | |
2723 | struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; | |
2724 | struct list_head *pos, *n; | |
2725 | struct ext4_li_request *elr; | |
2726 | unsigned long next_wakeup; | |
2727 | DEFINE_WAIT(wait); | |
bfff6873 LC |
2728 | |
2729 | BUG_ON(NULL == eli); | |
2730 | ||
2731 | eli->li_timer.data = (unsigned long)current; | |
2732 | eli->li_timer.function = ext4_lazyinode_timeout; | |
2733 | ||
2734 | eli->li_task = current; | |
2735 | wake_up(&eli->li_wait_task); | |
2736 | ||
2737 | cont_thread: | |
2738 | while (true) { | |
2739 | next_wakeup = MAX_JIFFY_OFFSET; | |
2740 | ||
2741 | mutex_lock(&eli->li_list_mtx); | |
2742 | if (list_empty(&eli->li_request_list)) { | |
2743 | mutex_unlock(&eli->li_list_mtx); | |
2744 | goto exit_thread; | |
2745 | } | |
2746 | ||
2747 | list_for_each_safe(pos, n, &eli->li_request_list) { | |
2748 | elr = list_entry(pos, struct ext4_li_request, | |
2749 | lr_request); | |
2750 | ||
b2c78cd0 TT |
2751 | if (time_after_eq(jiffies, elr->lr_next_sched)) { |
2752 | if (ext4_run_li_request(elr) != 0) { | |
2753 | /* error, remove the lazy_init job */ | |
2754 | ext4_remove_li_request(elr); | |
2755 | continue; | |
2756 | } | |
bfff6873 LC |
2757 | } |
2758 | ||
2759 | if (time_before(elr->lr_next_sched, next_wakeup)) | |
2760 | next_wakeup = elr->lr_next_sched; | |
2761 | } | |
2762 | mutex_unlock(&eli->li_list_mtx); | |
2763 | ||
2764 | if (freezing(current)) | |
2765 | refrigerator(); | |
2766 | ||
f4245bd4 LC |
2767 | if ((time_after_eq(jiffies, next_wakeup)) || |
2768 | (MAX_JIFFY_OFFSET == next_wakeup)) { | |
bfff6873 LC |
2769 | cond_resched(); |
2770 | continue; | |
2771 | } | |
2772 | ||
2773 | eli->li_timer.expires = next_wakeup; | |
2774 | add_timer(&eli->li_timer); | |
2775 | prepare_to_wait(&eli->li_wait_daemon, &wait, | |
2776 | TASK_INTERRUPTIBLE); | |
2777 | if (time_before(jiffies, next_wakeup)) | |
2778 | schedule(); | |
2779 | finish_wait(&eli->li_wait_daemon, &wait); | |
2780 | } | |
2781 | ||
2782 | exit_thread: | |
2783 | /* | |
2784 | * It looks like the request list is empty, but we need | |
2785 | * to check it under the li_list_mtx lock, to prevent any | |
2786 | * additions into it, and of course we should lock ext4_li_mtx | |
2787 | * to atomically free the list and ext4_li_info, because at | |
2788 | * this point another ext4 filesystem could be registering | |
2789 | * new one. | |
2790 | */ | |
2791 | mutex_lock(&ext4_li_mtx); | |
2792 | mutex_lock(&eli->li_list_mtx); | |
2793 | if (!list_empty(&eli->li_request_list)) { | |
2794 | mutex_unlock(&eli->li_list_mtx); | |
2795 | mutex_unlock(&ext4_li_mtx); | |
2796 | goto cont_thread; | |
2797 | } | |
2798 | mutex_unlock(&eli->li_list_mtx); | |
2799 | del_timer_sync(&ext4_li_info->li_timer); | |
2800 | eli->li_task = NULL; | |
2801 | wake_up(&eli->li_wait_task); | |
2802 | ||
2803 | kfree(ext4_li_info); | |
2804 | ext4_li_info = NULL; | |
2805 | mutex_unlock(&ext4_li_mtx); | |
2806 | ||
2807 | return 0; | |
2808 | } | |
2809 | ||
2810 | static void ext4_clear_request_list(void) | |
2811 | { | |
2812 | struct list_head *pos, *n; | |
2813 | struct ext4_li_request *elr; | |
2814 | ||
2815 | mutex_lock(&ext4_li_info->li_list_mtx); | |
bfff6873 LC |
2816 | list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { |
2817 | elr = list_entry(pos, struct ext4_li_request, | |
2818 | lr_request); | |
2819 | ext4_remove_li_request(elr); | |
2820 | } | |
2821 | mutex_unlock(&ext4_li_info->li_list_mtx); | |
2822 | } | |
2823 | ||
2824 | static int ext4_run_lazyinit_thread(void) | |
2825 | { | |
2826 | struct task_struct *t; | |
2827 | ||
2828 | t = kthread_run(ext4_lazyinit_thread, ext4_li_info, "ext4lazyinit"); | |
2829 | if (IS_ERR(t)) { | |
2830 | int err = PTR_ERR(t); | |
2831 | ext4_clear_request_list(); | |
2832 | del_timer_sync(&ext4_li_info->li_timer); | |
2833 | kfree(ext4_li_info); | |
2834 | ext4_li_info = NULL; | |
2835 | printk(KERN_CRIT "EXT4: error %d creating inode table " | |
2836 | "initialization thread\n", | |
2837 | err); | |
2838 | return err; | |
2839 | } | |
2840 | ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; | |
2841 | ||
2842 | wait_event(ext4_li_info->li_wait_task, ext4_li_info->li_task != NULL); | |
2843 | return 0; | |
2844 | } | |
2845 | ||
2846 | /* | |
2847 | * Check whether it make sense to run itable init. thread or not. | |
2848 | * If there is at least one uninitialized inode table, return | |
2849 | * corresponding group number, else the loop goes through all | |
2850 | * groups and return total number of groups. | |
2851 | */ | |
2852 | static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) | |
2853 | { | |
2854 | ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; | |
2855 | struct ext4_group_desc *gdp = NULL; | |
2856 | ||
2857 | for (group = 0; group < ngroups; group++) { | |
2858 | gdp = ext4_get_group_desc(sb, group, NULL); | |
2859 | if (!gdp) | |
2860 | continue; | |
2861 | ||
2862 | if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) | |
2863 | break; | |
2864 | } | |
2865 | ||
2866 | return group; | |
2867 | } | |
2868 | ||
2869 | static int ext4_li_info_new(void) | |
2870 | { | |
2871 | struct ext4_lazy_init *eli = NULL; | |
2872 | ||
2873 | eli = kzalloc(sizeof(*eli), GFP_KERNEL); | |
2874 | if (!eli) | |
2875 | return -ENOMEM; | |
2876 | ||
2877 | eli->li_task = NULL; | |
2878 | INIT_LIST_HEAD(&eli->li_request_list); | |
2879 | mutex_init(&eli->li_list_mtx); | |
2880 | ||
2881 | init_waitqueue_head(&eli->li_wait_daemon); | |
2882 | init_waitqueue_head(&eli->li_wait_task); | |
2883 | init_timer(&eli->li_timer); | |
2884 | eli->li_state |= EXT4_LAZYINIT_QUIT; | |
2885 | ||
2886 | ext4_li_info = eli; | |
2887 | ||
2888 | return 0; | |
2889 | } | |
2890 | ||
2891 | static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, | |
2892 | ext4_group_t start) | |
2893 | { | |
2894 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
2895 | struct ext4_li_request *elr; | |
2896 | unsigned long rnd; | |
2897 | ||
2898 | elr = kzalloc(sizeof(*elr), GFP_KERNEL); | |
2899 | if (!elr) | |
2900 | return NULL; | |
2901 | ||
2902 | elr->lr_super = sb; | |
2903 | elr->lr_sbi = sbi; | |
2904 | elr->lr_next_group = start; | |
2905 | ||
2906 | /* | |
2907 | * Randomize first schedule time of the request to | |
2908 | * spread the inode table initialization requests | |
2909 | * better. | |
2910 | */ | |
2911 | get_random_bytes(&rnd, sizeof(rnd)); | |
2912 | elr->lr_next_sched = jiffies + (unsigned long)rnd % | |
2913 | (EXT4_DEF_LI_MAX_START_DELAY * HZ); | |
2914 | ||
2915 | return elr; | |
2916 | } | |
2917 | ||
2918 | static int ext4_register_li_request(struct super_block *sb, | |
2919 | ext4_group_t first_not_zeroed) | |
2920 | { | |
2921 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
2922 | struct ext4_li_request *elr; | |
2923 | ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; | |
beed5ecb | 2924 | int ret; |
bfff6873 LC |
2925 | |
2926 | if (sbi->s_li_request != NULL) | |
beed5ecb | 2927 | return 0; |
bfff6873 LC |
2928 | |
2929 | if (first_not_zeroed == ngroups || | |
2930 | (sb->s_flags & MS_RDONLY) || | |
2931 | !test_opt(sb, INIT_INODE_TABLE)) { | |
2932 | sbi->s_li_request = NULL; | |
beed5ecb | 2933 | return 0; |
bfff6873 LC |
2934 | } |
2935 | ||
2936 | if (first_not_zeroed == ngroups) { | |
2937 | sbi->s_li_request = NULL; | |
beed5ecb | 2938 | return 0; |
bfff6873 LC |
2939 | } |
2940 | ||
2941 | elr = ext4_li_request_new(sb, first_not_zeroed); | |
beed5ecb NK |
2942 | if (!elr) |
2943 | return -ENOMEM; | |
bfff6873 LC |
2944 | |
2945 | mutex_lock(&ext4_li_mtx); | |
2946 | ||
2947 | if (NULL == ext4_li_info) { | |
2948 | ret = ext4_li_info_new(); | |
2949 | if (ret) | |
2950 | goto out; | |
2951 | } | |
2952 | ||
2953 | mutex_lock(&ext4_li_info->li_list_mtx); | |
2954 | list_add(&elr->lr_request, &ext4_li_info->li_request_list); | |
2955 | mutex_unlock(&ext4_li_info->li_list_mtx); | |
2956 | ||
2957 | sbi->s_li_request = elr; | |
2958 | ||
2959 | if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { | |
2960 | ret = ext4_run_lazyinit_thread(); | |
2961 | if (ret) | |
2962 | goto out; | |
2963 | } | |
bfff6873 | 2964 | out: |
beed5ecb NK |
2965 | mutex_unlock(&ext4_li_mtx); |
2966 | if (ret) | |
bfff6873 | 2967 | kfree(elr); |
bfff6873 LC |
2968 | return ret; |
2969 | } | |
2970 | ||
2971 | /* | |
2972 | * We do not need to lock anything since this is called on | |
2973 | * module unload. | |
2974 | */ | |
2975 | static void ext4_destroy_lazyinit_thread(void) | |
2976 | { | |
2977 | /* | |
2978 | * If thread exited earlier | |
2979 | * there's nothing to be done. | |
2980 | */ | |
2981 | if (!ext4_li_info) | |
2982 | return; | |
2983 | ||
2984 | ext4_clear_request_list(); | |
2985 | ||
2986 | while (ext4_li_info->li_task) { | |
2987 | wake_up(&ext4_li_info->li_wait_daemon); | |
2988 | wait_event(ext4_li_info->li_wait_task, | |
2989 | ext4_li_info->li_task == NULL); | |
2990 | } | |
2991 | } | |
2992 | ||
2b2d6d01 | 2993 | static int ext4_fill_super(struct super_block *sb, void *data, int silent) |
7477827f AK |
2994 | __releases(kernel_lock) |
2995 | __acquires(kernel_lock) | |
ac27a0ec | 2996 | { |
d4c402d9 | 2997 | char *orig_data = kstrdup(data, GFP_KERNEL); |
2b2d6d01 | 2998 | struct buffer_head *bh; |
617ba13b MC |
2999 | struct ext4_super_block *es = NULL; |
3000 | struct ext4_sb_info *sbi; | |
3001 | ext4_fsblk_t block; | |
3002 | ext4_fsblk_t sb_block = get_sb_block(&data); | |
70bbb3e0 | 3003 | ext4_fsblk_t logical_sb_block; |
ac27a0ec | 3004 | unsigned long offset = 0; |
ac27a0ec DK |
3005 | unsigned long journal_devnum = 0; |
3006 | unsigned long def_mount_opts; | |
3007 | struct inode *root; | |
9f6200bb | 3008 | char *cp; |
0390131b | 3009 | const char *descr; |
dcc7dae3 | 3010 | int ret = -ENOMEM; |
ac27a0ec | 3011 | int blocksize; |
4ec11028 TT |
3012 | unsigned int db_count; |
3013 | unsigned int i; | |
f287a1a5 | 3014 | int needs_recovery, has_huge_files; |
bd81d8ee | 3015 | __u64 blocks_count; |
833f4077 | 3016 | int err; |
b3881f74 | 3017 | unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; |
bfff6873 | 3018 | ext4_group_t first_not_zeroed; |
ac27a0ec DK |
3019 | |
3020 | sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); | |
3021 | if (!sbi) | |
dcc7dae3 | 3022 | goto out_free_orig; |
705895b6 PE |
3023 | |
3024 | sbi->s_blockgroup_lock = | |
3025 | kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); | |
3026 | if (!sbi->s_blockgroup_lock) { | |
3027 | kfree(sbi); | |
dcc7dae3 | 3028 | goto out_free_orig; |
705895b6 | 3029 | } |
ac27a0ec DK |
3030 | sb->s_fs_info = sbi; |
3031 | sbi->s_mount_opt = 0; | |
617ba13b MC |
3032 | sbi->s_resuid = EXT4_DEF_RESUID; |
3033 | sbi->s_resgid = EXT4_DEF_RESGID; | |
240799cd | 3034 | sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; |
d9c9bef1 | 3035 | sbi->s_sb_block = sb_block; |
f613dfcb TT |
3036 | if (sb->s_bdev->bd_part) |
3037 | sbi->s_sectors_written_start = | |
3038 | part_stat_read(sb->s_bdev->bd_part, sectors[1]); | |
ac27a0ec | 3039 | |
9f6200bb TT |
3040 | /* Cleanup superblock name */ |
3041 | for (cp = sb->s_id; (cp = strchr(cp, '/'));) | |
3042 | *cp = '!'; | |
3043 | ||
dcc7dae3 | 3044 | ret = -EINVAL; |
617ba13b | 3045 | blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); |
ac27a0ec | 3046 | if (!blocksize) { |
b31e1552 | 3047 | ext4_msg(sb, KERN_ERR, "unable to set blocksize"); |
ac27a0ec DK |
3048 | goto out_fail; |
3049 | } | |
3050 | ||
3051 | /* | |
617ba13b | 3052 | * The ext4 superblock will not be buffer aligned for other than 1kB |
ac27a0ec DK |
3053 | * block sizes. We need to calculate the offset from buffer start. |
3054 | */ | |
617ba13b | 3055 | if (blocksize != EXT4_MIN_BLOCK_SIZE) { |
70bbb3e0 AM |
3056 | logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; |
3057 | offset = do_div(logical_sb_block, blocksize); | |
ac27a0ec | 3058 | } else { |
70bbb3e0 | 3059 | logical_sb_block = sb_block; |
ac27a0ec DK |
3060 | } |
3061 | ||
70bbb3e0 | 3062 | if (!(bh = sb_bread(sb, logical_sb_block))) { |
b31e1552 | 3063 | ext4_msg(sb, KERN_ERR, "unable to read superblock"); |
ac27a0ec DK |
3064 | goto out_fail; |
3065 | } | |
3066 | /* | |
3067 | * Note: s_es must be initialized as soon as possible because | |
617ba13b | 3068 | * some ext4 macro-instructions depend on its value |
ac27a0ec | 3069 | */ |
617ba13b | 3070 | es = (struct ext4_super_block *) (((char *)bh->b_data) + offset); |
ac27a0ec DK |
3071 | sbi->s_es = es; |
3072 | sb->s_magic = le16_to_cpu(es->s_magic); | |
617ba13b MC |
3073 | if (sb->s_magic != EXT4_SUPER_MAGIC) |
3074 | goto cantfind_ext4; | |
afc32f7e | 3075 | sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); |
ac27a0ec DK |
3076 | |
3077 | /* Set defaults before we parse the mount options */ | |
3078 | def_mount_opts = le32_to_cpu(es->s_default_mount_opts); | |
fd8c37ec | 3079 | set_opt(sb, INIT_INODE_TABLE); |
617ba13b | 3080 | if (def_mount_opts & EXT4_DEFM_DEBUG) |
fd8c37ec | 3081 | set_opt(sb, DEBUG); |
437ca0fd DM |
3082 | if (def_mount_opts & EXT4_DEFM_BSDGROUPS) { |
3083 | ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups", | |
3084 | "2.6.38"); | |
fd8c37ec | 3085 | set_opt(sb, GRPID); |
437ca0fd | 3086 | } |
617ba13b | 3087 | if (def_mount_opts & EXT4_DEFM_UID16) |
fd8c37ec | 3088 | set_opt(sb, NO_UID32); |
03010a33 | 3089 | #ifdef CONFIG_EXT4_FS_XATTR |
617ba13b | 3090 | if (def_mount_opts & EXT4_DEFM_XATTR_USER) |
fd8c37ec | 3091 | set_opt(sb, XATTR_USER); |
2e7842b8 | 3092 | #endif |
03010a33 | 3093 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
617ba13b | 3094 | if (def_mount_opts & EXT4_DEFM_ACL) |
fd8c37ec | 3095 | set_opt(sb, POSIX_ACL); |
2e7842b8 | 3096 | #endif |
617ba13b | 3097 | if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) |
fd8c37ec | 3098 | set_opt(sb, JOURNAL_DATA); |
617ba13b | 3099 | else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) |
fd8c37ec | 3100 | set_opt(sb, ORDERED_DATA); |
617ba13b | 3101 | else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) |
fd8c37ec | 3102 | set_opt(sb, WRITEBACK_DATA); |
617ba13b MC |
3103 | |
3104 | if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) | |
fd8c37ec | 3105 | set_opt(sb, ERRORS_PANIC); |
bb4f397a | 3106 | else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) |
fd8c37ec | 3107 | set_opt(sb, ERRORS_CONT); |
bb4f397a | 3108 | else |
fd8c37ec | 3109 | set_opt(sb, ERRORS_RO); |
8b67f04a | 3110 | if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY) |
fd8c37ec | 3111 | set_opt(sb, BLOCK_VALIDITY); |
8b67f04a | 3112 | if (def_mount_opts & EXT4_DEFM_DISCARD) |
fd8c37ec | 3113 | set_opt(sb, DISCARD); |
ac27a0ec DK |
3114 | |
3115 | sbi->s_resuid = le16_to_cpu(es->s_def_resuid); | |
3116 | sbi->s_resgid = le16_to_cpu(es->s_def_resgid); | |
30773840 TT |
3117 | sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; |
3118 | sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; | |
3119 | sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; | |
ac27a0ec | 3120 | |
8b67f04a | 3121 | if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) |
fd8c37ec | 3122 | set_opt(sb, BARRIER); |
ac27a0ec | 3123 | |
dd919b98 AK |
3124 | /* |
3125 | * enable delayed allocation by default | |
3126 | * Use -o nodelalloc to turn it off | |
3127 | */ | |
8b67f04a TT |
3128 | if (!IS_EXT3_SB(sb) && |
3129 | ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) | |
fd8c37ec | 3130 | set_opt(sb, DELALLOC); |
dd919b98 | 3131 | |
8b67f04a TT |
3132 | if (!parse_options((char *) sbi->s_es->s_mount_opts, sb, |
3133 | &journal_devnum, &journal_ioprio, NULL, 0)) { | |
3134 | ext4_msg(sb, KERN_WARNING, | |
3135 | "failed to parse options in superblock: %s", | |
3136 | sbi->s_es->s_mount_opts); | |
3137 | } | |
b3881f74 TT |
3138 | if (!parse_options((char *) data, sb, &journal_devnum, |
3139 | &journal_ioprio, NULL, 0)) | |
ac27a0ec DK |
3140 | goto failed_mount; |
3141 | ||
3142 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
482a7425 | 3143 | (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); |
ac27a0ec | 3144 | |
617ba13b MC |
3145 | if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && |
3146 | (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) || | |
3147 | EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) || | |
3148 | EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U))) | |
b31e1552 ES |
3149 | ext4_msg(sb, KERN_WARNING, |
3150 | "feature flags set on rev 0 fs, " | |
3151 | "running e2fsck is recommended"); | |
469108ff | 3152 | |
ac27a0ec DK |
3153 | /* |
3154 | * Check feature flags regardless of the revision level, since we | |
3155 | * previously didn't change the revision level when setting the flags, | |
3156 | * so there is a chance incompat flags are set on a rev 0 filesystem. | |
3157 | */ | |
a13fb1a4 | 3158 | if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY))) |
ac27a0ec | 3159 | goto failed_mount; |
a13fb1a4 | 3160 | |
ac27a0ec DK |
3161 | blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); |
3162 | ||
617ba13b MC |
3163 | if (blocksize < EXT4_MIN_BLOCK_SIZE || |
3164 | blocksize > EXT4_MAX_BLOCK_SIZE) { | |
b31e1552 ES |
3165 | ext4_msg(sb, KERN_ERR, |
3166 | "Unsupported filesystem blocksize %d", blocksize); | |
ac27a0ec DK |
3167 | goto failed_mount; |
3168 | } | |
3169 | ||
ac27a0ec | 3170 | if (sb->s_blocksize != blocksize) { |
ce40733c AK |
3171 | /* Validate the filesystem blocksize */ |
3172 | if (!sb_set_blocksize(sb, blocksize)) { | |
b31e1552 | 3173 | ext4_msg(sb, KERN_ERR, "bad block size %d", |
ce40733c | 3174 | blocksize); |
ac27a0ec DK |
3175 | goto failed_mount; |
3176 | } | |
3177 | ||
2b2d6d01 | 3178 | brelse(bh); |
70bbb3e0 AM |
3179 | logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; |
3180 | offset = do_div(logical_sb_block, blocksize); | |
3181 | bh = sb_bread(sb, logical_sb_block); | |
ac27a0ec | 3182 | if (!bh) { |
b31e1552 ES |
3183 | ext4_msg(sb, KERN_ERR, |
3184 | "Can't read superblock on 2nd try"); | |
ac27a0ec DK |
3185 | goto failed_mount; |
3186 | } | |
617ba13b | 3187 | es = (struct ext4_super_block *)(((char *)bh->b_data) + offset); |
ac27a0ec | 3188 | sbi->s_es = es; |
617ba13b | 3189 | if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { |
b31e1552 ES |
3190 | ext4_msg(sb, KERN_ERR, |
3191 | "Magic mismatch, very weird!"); | |
ac27a0ec DK |
3192 | goto failed_mount; |
3193 | } | |
3194 | } | |
3195 | ||
a13fb1a4 ES |
3196 | has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb, |
3197 | EXT4_FEATURE_RO_COMPAT_HUGE_FILE); | |
f287a1a5 TT |
3198 | sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, |
3199 | has_huge_files); | |
3200 | sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); | |
ac27a0ec | 3201 | |
617ba13b MC |
3202 | if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { |
3203 | sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; | |
3204 | sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; | |
ac27a0ec DK |
3205 | } else { |
3206 | sbi->s_inode_size = le16_to_cpu(es->s_inode_size); | |
3207 | sbi->s_first_ino = le32_to_cpu(es->s_first_ino); | |
617ba13b | 3208 | if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || |
1330593e | 3209 | (!is_power_of_2(sbi->s_inode_size)) || |
ac27a0ec | 3210 | (sbi->s_inode_size > blocksize)) { |
b31e1552 ES |
3211 | ext4_msg(sb, KERN_ERR, |
3212 | "unsupported inode size: %d", | |
2b2d6d01 | 3213 | sbi->s_inode_size); |
ac27a0ec DK |
3214 | goto failed_mount; |
3215 | } | |
ef7f3835 KS |
3216 | if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) |
3217 | sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2); | |
ac27a0ec | 3218 | } |
0b8e58a1 | 3219 | |
0d1ee42f AR |
3220 | sbi->s_desc_size = le16_to_cpu(es->s_desc_size); |
3221 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) { | |
8fadc143 | 3222 | if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || |
0d1ee42f | 3223 | sbi->s_desc_size > EXT4_MAX_DESC_SIZE || |
d8ea6cf8 | 3224 | !is_power_of_2(sbi->s_desc_size)) { |
b31e1552 ES |
3225 | ext4_msg(sb, KERN_ERR, |
3226 | "unsupported descriptor size %lu", | |
0d1ee42f AR |
3227 | sbi->s_desc_size); |
3228 | goto failed_mount; | |
3229 | } | |
3230 | } else | |
3231 | sbi->s_desc_size = EXT4_MIN_DESC_SIZE; | |
0b8e58a1 | 3232 | |
ac27a0ec | 3233 | sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); |
ac27a0ec | 3234 | sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); |
b47b6f38 | 3235 | if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0) |
617ba13b | 3236 | goto cantfind_ext4; |
0b8e58a1 | 3237 | |
617ba13b | 3238 | sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); |
ac27a0ec | 3239 | if (sbi->s_inodes_per_block == 0) |
617ba13b | 3240 | goto cantfind_ext4; |
ac27a0ec DK |
3241 | sbi->s_itb_per_group = sbi->s_inodes_per_group / |
3242 | sbi->s_inodes_per_block; | |
0d1ee42f | 3243 | sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); |
ac27a0ec DK |
3244 | sbi->s_sbh = bh; |
3245 | sbi->s_mount_state = le16_to_cpu(es->s_state); | |
e57aa839 FW |
3246 | sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); |
3247 | sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); | |
0b8e58a1 | 3248 | |
2b2d6d01 | 3249 | for (i = 0; i < 4; i++) |
ac27a0ec DK |
3250 | sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); |
3251 | sbi->s_def_hash_version = es->s_def_hash_version; | |
f99b2589 TT |
3252 | i = le32_to_cpu(es->s_flags); |
3253 | if (i & EXT2_FLAGS_UNSIGNED_HASH) | |
3254 | sbi->s_hash_unsigned = 3; | |
3255 | else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { | |
3256 | #ifdef __CHAR_UNSIGNED__ | |
3257 | es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); | |
3258 | sbi->s_hash_unsigned = 3; | |
3259 | #else | |
3260 | es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); | |
3261 | #endif | |
3262 | sb->s_dirt = 1; | |
3263 | } | |
ac27a0ec DK |
3264 | |
3265 | if (sbi->s_blocks_per_group > blocksize * 8) { | |
b31e1552 ES |
3266 | ext4_msg(sb, KERN_ERR, |
3267 | "#blocks per group too big: %lu", | |
2b2d6d01 | 3268 | sbi->s_blocks_per_group); |
ac27a0ec DK |
3269 | goto failed_mount; |
3270 | } | |
ac27a0ec | 3271 | if (sbi->s_inodes_per_group > blocksize * 8) { |
b31e1552 ES |
3272 | ext4_msg(sb, KERN_ERR, |
3273 | "#inodes per group too big: %lu", | |
2b2d6d01 | 3274 | sbi->s_inodes_per_group); |
ac27a0ec DK |
3275 | goto failed_mount; |
3276 | } | |
3277 | ||
bf43d84b ES |
3278 | /* |
3279 | * Test whether we have more sectors than will fit in sector_t, | |
3280 | * and whether the max offset is addressable by the page cache. | |
3281 | */ | |
5a9ae68a | 3282 | err = generic_check_addressable(sb->s_blocksize_bits, |
30ca22c7 | 3283 | ext4_blocks_count(es)); |
5a9ae68a | 3284 | if (err) { |
b31e1552 | 3285 | ext4_msg(sb, KERN_ERR, "filesystem" |
bf43d84b | 3286 | " too large to mount safely on this system"); |
ac27a0ec | 3287 | if (sizeof(sector_t) < 8) |
90c699a9 | 3288 | ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled"); |
5a9ae68a | 3289 | ret = err; |
ac27a0ec DK |
3290 | goto failed_mount; |
3291 | } | |
3292 | ||
617ba13b MC |
3293 | if (EXT4_BLOCKS_PER_GROUP(sb) == 0) |
3294 | goto cantfind_ext4; | |
e7c95593 | 3295 | |
0f2ddca6 FTN |
3296 | /* check blocks count against device size */ |
3297 | blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; | |
3298 | if (blocks_count && ext4_blocks_count(es) > blocks_count) { | |
b31e1552 ES |
3299 | ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " |
3300 | "exceeds size of device (%llu blocks)", | |
0f2ddca6 FTN |
3301 | ext4_blocks_count(es), blocks_count); |
3302 | goto failed_mount; | |
3303 | } | |
3304 | ||
0b8e58a1 AD |
3305 | /* |
3306 | * It makes no sense for the first data block to be beyond the end | |
3307 | * of the filesystem. | |
3308 | */ | |
3309 | if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { | |
b31e1552 ES |
3310 | ext4_msg(sb, KERN_WARNING, "bad geometry: first data" |
3311 | "block %u is beyond end of filesystem (%llu)", | |
3312 | le32_to_cpu(es->s_first_data_block), | |
3313 | ext4_blocks_count(es)); | |
e7c95593 ES |
3314 | goto failed_mount; |
3315 | } | |
bd81d8ee LV |
3316 | blocks_count = (ext4_blocks_count(es) - |
3317 | le32_to_cpu(es->s_first_data_block) + | |
3318 | EXT4_BLOCKS_PER_GROUP(sb) - 1); | |
3319 | do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); | |
4ec11028 | 3320 | if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { |
b31e1552 | 3321 | ext4_msg(sb, KERN_WARNING, "groups count too large: %u " |
4ec11028 | 3322 | "(block count %llu, first data block %u, " |
b31e1552 | 3323 | "blocks per group %lu)", sbi->s_groups_count, |
4ec11028 TT |
3324 | ext4_blocks_count(es), |
3325 | le32_to_cpu(es->s_first_data_block), | |
3326 | EXT4_BLOCKS_PER_GROUP(sb)); | |
3327 | goto failed_mount; | |
3328 | } | |
bd81d8ee | 3329 | sbi->s_groups_count = blocks_count; |
fb0a387d ES |
3330 | sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, |
3331 | (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); | |
617ba13b MC |
3332 | db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / |
3333 | EXT4_DESC_PER_BLOCK(sb); | |
2b2d6d01 | 3334 | sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *), |
ac27a0ec DK |
3335 | GFP_KERNEL); |
3336 | if (sbi->s_group_desc == NULL) { | |
b31e1552 | 3337 | ext4_msg(sb, KERN_ERR, "not enough memory"); |
ac27a0ec DK |
3338 | goto failed_mount; |
3339 | } | |
3340 | ||
3244fcb1 | 3341 | #ifdef CONFIG_PROC_FS |
9f6200bb TT |
3342 | if (ext4_proc_root) |
3343 | sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root); | |
3244fcb1 | 3344 | #endif |
240799cd | 3345 | |
705895b6 | 3346 | bgl_lock_init(sbi->s_blockgroup_lock); |
ac27a0ec DK |
3347 | |
3348 | for (i = 0; i < db_count; i++) { | |
70bbb3e0 | 3349 | block = descriptor_loc(sb, logical_sb_block, i); |
ac27a0ec DK |
3350 | sbi->s_group_desc[i] = sb_bread(sb, block); |
3351 | if (!sbi->s_group_desc[i]) { | |
b31e1552 ES |
3352 | ext4_msg(sb, KERN_ERR, |
3353 | "can't read group descriptor %d", i); | |
ac27a0ec DK |
3354 | db_count = i; |
3355 | goto failed_mount2; | |
3356 | } | |
3357 | } | |
bfff6873 | 3358 | if (!ext4_check_descriptors(sb, &first_not_zeroed)) { |
b31e1552 | 3359 | ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); |
ac27a0ec DK |
3360 | goto failed_mount2; |
3361 | } | |
772cb7c8 JS |
3362 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) |
3363 | if (!ext4_fill_flex_info(sb)) { | |
b31e1552 ES |
3364 | ext4_msg(sb, KERN_ERR, |
3365 | "unable to initialize " | |
3366 | "flex_bg meta info!"); | |
772cb7c8 JS |
3367 | goto failed_mount2; |
3368 | } | |
3369 | ||
ac27a0ec DK |
3370 | sbi->s_gdb_count = db_count; |
3371 | get_random_bytes(&sbi->s_next_generation, sizeof(u32)); | |
3372 | spin_lock_init(&sbi->s_next_gen_lock); | |
3373 | ||
ce7e010a TT |
3374 | err = percpu_counter_init(&sbi->s_freeblocks_counter, |
3375 | ext4_count_free_blocks(sb)); | |
3376 | if (!err) { | |
3377 | err = percpu_counter_init(&sbi->s_freeinodes_counter, | |
3378 | ext4_count_free_inodes(sb)); | |
3379 | } | |
3380 | if (!err) { | |
3381 | err = percpu_counter_init(&sbi->s_dirs_counter, | |
3382 | ext4_count_dirs(sb)); | |
3383 | } | |
3384 | if (!err) { | |
3385 | err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0); | |
3386 | } | |
3387 | if (err) { | |
3388 | ext4_msg(sb, KERN_ERR, "insufficient memory"); | |
3389 | goto failed_mount3; | |
3390 | } | |
3391 | ||
c9de560d | 3392 | sbi->s_stripe = ext4_get_stripe_size(sbi); |
55138e0b | 3393 | sbi->s_max_writeback_mb_bump = 128; |
c9de560d | 3394 | |
ac27a0ec DK |
3395 | /* |
3396 | * set up enough so that it can read an inode | |
3397 | */ | |
9ca92389 TT |
3398 | if (!test_opt(sb, NOLOAD) && |
3399 | EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) | |
3400 | sb->s_op = &ext4_sops; | |
3401 | else | |
3402 | sb->s_op = &ext4_nojournal_sops; | |
617ba13b MC |
3403 | sb->s_export_op = &ext4_export_ops; |
3404 | sb->s_xattr = ext4_xattr_handlers; | |
ac27a0ec | 3405 | #ifdef CONFIG_QUOTA |
617ba13b MC |
3406 | sb->s_qcop = &ext4_qctl_operations; |
3407 | sb->dq_op = &ext4_quota_operations; | |
ac27a0ec DK |
3408 | #endif |
3409 | INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ | |
3b9d4ed2 | 3410 | mutex_init(&sbi->s_orphan_lock); |
32ed5058 | 3411 | mutex_init(&sbi->s_resize_lock); |
ac27a0ec DK |
3412 | |
3413 | sb->s_root = NULL; | |
3414 | ||
3415 | needs_recovery = (es->s_last_orphan != 0 || | |
617ba13b MC |
3416 | EXT4_HAS_INCOMPAT_FEATURE(sb, |
3417 | EXT4_FEATURE_INCOMPAT_RECOVER)); | |
ac27a0ec DK |
3418 | |
3419 | /* | |
3420 | * The first inode we look at is the journal inode. Don't try | |
3421 | * root first: it may be modified in the journal! | |
3422 | */ | |
3423 | if (!test_opt(sb, NOLOAD) && | |
617ba13b MC |
3424 | EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { |
3425 | if (ext4_load_journal(sb, es, journal_devnum)) | |
ac27a0ec | 3426 | goto failed_mount3; |
0390131b FM |
3427 | } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) && |
3428 | EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { | |
b31e1552 ES |
3429 | ext4_msg(sb, KERN_ERR, "required journal recovery " |
3430 | "suppressed and not mounted read-only"); | |
744692dc | 3431 | goto failed_mount_wq; |
ac27a0ec | 3432 | } else { |
fd8c37ec TT |
3433 | clear_opt(sb, DATA_FLAGS); |
3434 | set_opt(sb, WRITEBACK_DATA); | |
0390131b FM |
3435 | sbi->s_journal = NULL; |
3436 | needs_recovery = 0; | |
3437 | goto no_journal; | |
ac27a0ec DK |
3438 | } |
3439 | ||
eb40a09c JS |
3440 | if (ext4_blocks_count(es) > 0xffffffffULL && |
3441 | !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, | |
3442 | JBD2_FEATURE_INCOMPAT_64BIT)) { | |
b31e1552 | 3443 | ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); |
744692dc | 3444 | goto failed_mount_wq; |
eb40a09c JS |
3445 | } |
3446 | ||
d4da6c9c LT |
3447 | if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { |
3448 | jbd2_journal_set_features(sbi->s_journal, | |
3449 | JBD2_FEATURE_COMPAT_CHECKSUM, 0, | |
818d276c | 3450 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); |
d4da6c9c LT |
3451 | } else if (test_opt(sb, JOURNAL_CHECKSUM)) { |
3452 | jbd2_journal_set_features(sbi->s_journal, | |
3453 | JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0); | |
818d276c GS |
3454 | jbd2_journal_clear_features(sbi->s_journal, 0, 0, |
3455 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); | |
d4da6c9c LT |
3456 | } else { |
3457 | jbd2_journal_clear_features(sbi->s_journal, | |
3458 | JBD2_FEATURE_COMPAT_CHECKSUM, 0, | |
3459 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); | |
3460 | } | |
818d276c | 3461 | |
ac27a0ec DK |
3462 | /* We have now updated the journal if required, so we can |
3463 | * validate the data journaling mode. */ | |
3464 | switch (test_opt(sb, DATA_FLAGS)) { | |
3465 | case 0: | |
3466 | /* No mode set, assume a default based on the journal | |
63f57933 AM |
3467 | * capabilities: ORDERED_DATA if the journal can |
3468 | * cope, else JOURNAL_DATA | |
3469 | */ | |
dab291af MC |
3470 | if (jbd2_journal_check_available_features |
3471 | (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) | |
fd8c37ec | 3472 | set_opt(sb, ORDERED_DATA); |
ac27a0ec | 3473 | else |
fd8c37ec | 3474 | set_opt(sb, JOURNAL_DATA); |
ac27a0ec DK |
3475 | break; |
3476 | ||
617ba13b MC |
3477 | case EXT4_MOUNT_ORDERED_DATA: |
3478 | case EXT4_MOUNT_WRITEBACK_DATA: | |
dab291af MC |
3479 | if (!jbd2_journal_check_available_features |
3480 | (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { | |
b31e1552 ES |
3481 | ext4_msg(sb, KERN_ERR, "Journal does not support " |
3482 | "requested data journaling mode"); | |
744692dc | 3483 | goto failed_mount_wq; |
ac27a0ec DK |
3484 | } |
3485 | default: | |
3486 | break; | |
3487 | } | |
b3881f74 | 3488 | set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); |
ac27a0ec | 3489 | |
ce7e010a TT |
3490 | /* |
3491 | * The journal may have updated the bg summary counts, so we | |
3492 | * need to update the global counters. | |
3493 | */ | |
3494 | percpu_counter_set(&sbi->s_freeblocks_counter, | |
3495 | ext4_count_free_blocks(sb)); | |
3496 | percpu_counter_set(&sbi->s_freeinodes_counter, | |
3497 | ext4_count_free_inodes(sb)); | |
3498 | percpu_counter_set(&sbi->s_dirs_counter, | |
3499 | ext4_count_dirs(sb)); | |
3500 | percpu_counter_set(&sbi->s_dirtyblocks_counter, 0); | |
206f7ab4 | 3501 | |
ce7e010a | 3502 | no_journal: |
4c0425ff MC |
3503 | EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten"); |
3504 | if (!EXT4_SB(sb)->dio_unwritten_wq) { | |
3505 | printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n"); | |
3506 | goto failed_mount_wq; | |
3507 | } | |
3508 | ||
ac27a0ec | 3509 | /* |
dab291af | 3510 | * The jbd2_journal_load will have done any necessary log recovery, |
ac27a0ec DK |
3511 | * so we can safely mount the rest of the filesystem now. |
3512 | */ | |
3513 | ||
1d1fe1ee DH |
3514 | root = ext4_iget(sb, EXT4_ROOT_INO); |
3515 | if (IS_ERR(root)) { | |
b31e1552 | 3516 | ext4_msg(sb, KERN_ERR, "get root inode failed"); |
1d1fe1ee | 3517 | ret = PTR_ERR(root); |
ac27a0ec DK |
3518 | goto failed_mount4; |
3519 | } | |
3520 | if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { | |
1d1fe1ee | 3521 | iput(root); |
b31e1552 | 3522 | ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); |
ac27a0ec DK |
3523 | goto failed_mount4; |
3524 | } | |
1d1fe1ee DH |
3525 | sb->s_root = d_alloc_root(root); |
3526 | if (!sb->s_root) { | |
b31e1552 | 3527 | ext4_msg(sb, KERN_ERR, "get root dentry failed"); |
1d1fe1ee DH |
3528 | iput(root); |
3529 | ret = -ENOMEM; | |
3530 | goto failed_mount4; | |
3531 | } | |
ac27a0ec | 3532 | |
2b2d6d01 | 3533 | ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY); |
ef7f3835 KS |
3534 | |
3535 | /* determine the minimum size of new large inodes, if present */ | |
3536 | if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { | |
3537 | sbi->s_want_extra_isize = sizeof(struct ext4_inode) - | |
3538 | EXT4_GOOD_OLD_INODE_SIZE; | |
3539 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, | |
3540 | EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) { | |
3541 | if (sbi->s_want_extra_isize < | |
3542 | le16_to_cpu(es->s_want_extra_isize)) | |
3543 | sbi->s_want_extra_isize = | |
3544 | le16_to_cpu(es->s_want_extra_isize); | |
3545 | if (sbi->s_want_extra_isize < | |
3546 | le16_to_cpu(es->s_min_extra_isize)) | |
3547 | sbi->s_want_extra_isize = | |
3548 | le16_to_cpu(es->s_min_extra_isize); | |
3549 | } | |
3550 | } | |
3551 | /* Check if enough inode space is available */ | |
3552 | if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize > | |
3553 | sbi->s_inode_size) { | |
3554 | sbi->s_want_extra_isize = sizeof(struct ext4_inode) - | |
3555 | EXT4_GOOD_OLD_INODE_SIZE; | |
b31e1552 ES |
3556 | ext4_msg(sb, KERN_INFO, "required extra inode space not" |
3557 | "available"); | |
ef7f3835 KS |
3558 | } |
3559 | ||
90576c0b TT |
3560 | if (test_opt(sb, DELALLOC) && |
3561 | (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) { | |
b31e1552 ES |
3562 | ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - " |
3563 | "requested data journaling mode"); | |
fd8c37ec | 3564 | clear_opt(sb, DELALLOC); |
90576c0b | 3565 | } |
744692dc JZ |
3566 | if (test_opt(sb, DIOREAD_NOLOCK)) { |
3567 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { | |
3568 | ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock " | |
3569 | "option - requested data journaling mode"); | |
fd8c37ec | 3570 | clear_opt(sb, DIOREAD_NOLOCK); |
744692dc JZ |
3571 | } |
3572 | if (sb->s_blocksize < PAGE_SIZE) { | |
3573 | ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock " | |
3574 | "option - block size is too small"); | |
fd8c37ec | 3575 | clear_opt(sb, DIOREAD_NOLOCK); |
744692dc JZ |
3576 | } |
3577 | } | |
c2774d84 | 3578 | |
6fd058f7 TT |
3579 | err = ext4_setup_system_zone(sb); |
3580 | if (err) { | |
b31e1552 | 3581 | ext4_msg(sb, KERN_ERR, "failed to initialize system " |
fbe845dd | 3582 | "zone (%d)", err); |
6fd058f7 TT |
3583 | goto failed_mount4; |
3584 | } | |
3585 | ||
c2774d84 AK |
3586 | ext4_ext_init(sb); |
3587 | err = ext4_mb_init(sb, needs_recovery); | |
3588 | if (err) { | |
421f91d2 | 3589 | ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", |
b31e1552 | 3590 | err); |
c2774d84 AK |
3591 | goto failed_mount4; |
3592 | } | |
3593 | ||
bfff6873 LC |
3594 | err = ext4_register_li_request(sb, first_not_zeroed); |
3595 | if (err) | |
3596 | goto failed_mount4; | |
3597 | ||
3197ebdb TT |
3598 | sbi->s_kobj.kset = ext4_kset; |
3599 | init_completion(&sbi->s_kobj_unregister); | |
3600 | err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL, | |
3601 | "%s", sb->s_id); | |
3602 | if (err) { | |
3603 | ext4_mb_release(sb); | |
3604 | ext4_ext_release(sb); | |
3605 | goto failed_mount4; | |
3606 | }; | |
3607 | ||
617ba13b MC |
3608 | EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; |
3609 | ext4_orphan_cleanup(sb, es); | |
3610 | EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; | |
0390131b | 3611 | if (needs_recovery) { |
b31e1552 | 3612 | ext4_msg(sb, KERN_INFO, "recovery complete"); |
0390131b FM |
3613 | ext4_mark_recovery_complete(sb, es); |
3614 | } | |
3615 | if (EXT4_SB(sb)->s_journal) { | |
3616 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) | |
3617 | descr = " journalled data mode"; | |
3618 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) | |
3619 | descr = " ordered data mode"; | |
3620 | else | |
3621 | descr = " writeback data mode"; | |
3622 | } else | |
3623 | descr = "out journal"; | |
3624 | ||
d4c402d9 | 3625 | ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " |
8b67f04a TT |
3626 | "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts, |
3627 | *sbi->s_es->s_mount_opts ? "; " : "", orig_data); | |
ac27a0ec | 3628 | |
66e61a9e TT |
3629 | init_timer(&sbi->s_err_report); |
3630 | sbi->s_err_report.function = print_daily_error_info; | |
3631 | sbi->s_err_report.data = (unsigned long) sb; | |
3632 | if (es->s_error_count) | |
3633 | mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ | |
ac27a0ec | 3634 | |
d4c402d9 | 3635 | kfree(orig_data); |
ac27a0ec DK |
3636 | return 0; |
3637 | ||
617ba13b | 3638 | cantfind_ext4: |
ac27a0ec | 3639 | if (!silent) |
b31e1552 | 3640 | ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); |
ac27a0ec DK |
3641 | goto failed_mount; |
3642 | ||
3643 | failed_mount4: | |
b31e1552 | 3644 | ext4_msg(sb, KERN_ERR, "mount failed"); |
4c0425ff MC |
3645 | destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq); |
3646 | failed_mount_wq: | |
6fd058f7 | 3647 | ext4_release_system_zone(sb); |
0390131b FM |
3648 | if (sbi->s_journal) { |
3649 | jbd2_journal_destroy(sbi->s_journal); | |
3650 | sbi->s_journal = NULL; | |
3651 | } | |
ac27a0ec | 3652 | failed_mount3: |
c5ca7c76 TT |
3653 | if (sbi->s_flex_groups) { |
3654 | if (is_vmalloc_addr(sbi->s_flex_groups)) | |
3655 | vfree(sbi->s_flex_groups); | |
3656 | else | |
3657 | kfree(sbi->s_flex_groups); | |
3658 | } | |
ce7e010a TT |
3659 | percpu_counter_destroy(&sbi->s_freeblocks_counter); |
3660 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
3661 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
3662 | percpu_counter_destroy(&sbi->s_dirtyblocks_counter); | |
ac27a0ec DK |
3663 | failed_mount2: |
3664 | for (i = 0; i < db_count; i++) | |
3665 | brelse(sbi->s_group_desc[i]); | |
3666 | kfree(sbi->s_group_desc); | |
3667 | failed_mount: | |
240799cd | 3668 | if (sbi->s_proc) { |
9f6200bb | 3669 | remove_proc_entry(sb->s_id, ext4_proc_root); |
240799cd | 3670 | } |
ac27a0ec DK |
3671 | #ifdef CONFIG_QUOTA |
3672 | for (i = 0; i < MAXQUOTAS; i++) | |
3673 | kfree(sbi->s_qf_names[i]); | |
3674 | #endif | |
617ba13b | 3675 | ext4_blkdev_remove(sbi); |
ac27a0ec DK |
3676 | brelse(bh); |
3677 | out_fail: | |
3678 | sb->s_fs_info = NULL; | |
f6830165 | 3679 | kfree(sbi->s_blockgroup_lock); |
ac27a0ec | 3680 | kfree(sbi); |
dcc7dae3 | 3681 | out_free_orig: |
d4c402d9 | 3682 | kfree(orig_data); |
1d1fe1ee | 3683 | return ret; |
ac27a0ec DK |
3684 | } |
3685 | ||
3686 | /* | |
3687 | * Setup any per-fs journal parameters now. We'll do this both on | |
3688 | * initial mount, once the journal has been initialised but before we've | |
3689 | * done any recovery; and again on any subsequent remount. | |
3690 | */ | |
617ba13b | 3691 | static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) |
ac27a0ec | 3692 | { |
617ba13b | 3693 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 3694 | |
30773840 TT |
3695 | journal->j_commit_interval = sbi->s_commit_interval; |
3696 | journal->j_min_batch_time = sbi->s_min_batch_time; | |
3697 | journal->j_max_batch_time = sbi->s_max_batch_time; | |
ac27a0ec | 3698 | |
a931da6a | 3699 | write_lock(&journal->j_state_lock); |
ac27a0ec | 3700 | if (test_opt(sb, BARRIER)) |
dab291af | 3701 | journal->j_flags |= JBD2_BARRIER; |
ac27a0ec | 3702 | else |
dab291af | 3703 | journal->j_flags &= ~JBD2_BARRIER; |
5bf5683a HK |
3704 | if (test_opt(sb, DATA_ERR_ABORT)) |
3705 | journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; | |
3706 | else | |
3707 | journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; | |
a931da6a | 3708 | write_unlock(&journal->j_state_lock); |
ac27a0ec DK |
3709 | } |
3710 | ||
617ba13b | 3711 | static journal_t *ext4_get_journal(struct super_block *sb, |
ac27a0ec DK |
3712 | unsigned int journal_inum) |
3713 | { | |
3714 | struct inode *journal_inode; | |
3715 | journal_t *journal; | |
3716 | ||
0390131b FM |
3717 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
3718 | ||
ac27a0ec DK |
3719 | /* First, test for the existence of a valid inode on disk. Bad |
3720 | * things happen if we iget() an unused inode, as the subsequent | |
3721 | * iput() will try to delete it. */ | |
3722 | ||
1d1fe1ee DH |
3723 | journal_inode = ext4_iget(sb, journal_inum); |
3724 | if (IS_ERR(journal_inode)) { | |
b31e1552 | 3725 | ext4_msg(sb, KERN_ERR, "no journal found"); |
ac27a0ec DK |
3726 | return NULL; |
3727 | } | |
3728 | if (!journal_inode->i_nlink) { | |
3729 | make_bad_inode(journal_inode); | |
3730 | iput(journal_inode); | |
b31e1552 | 3731 | ext4_msg(sb, KERN_ERR, "journal inode is deleted"); |
ac27a0ec DK |
3732 | return NULL; |
3733 | } | |
3734 | ||
e5f8eab8 | 3735 | jbd_debug(2, "Journal inode found at %p: %lld bytes\n", |
ac27a0ec | 3736 | journal_inode, journal_inode->i_size); |
1d1fe1ee | 3737 | if (!S_ISREG(journal_inode->i_mode)) { |
b31e1552 | 3738 | ext4_msg(sb, KERN_ERR, "invalid journal inode"); |
ac27a0ec DK |
3739 | iput(journal_inode); |
3740 | return NULL; | |
3741 | } | |
3742 | ||
dab291af | 3743 | journal = jbd2_journal_init_inode(journal_inode); |
ac27a0ec | 3744 | if (!journal) { |
b31e1552 | 3745 | ext4_msg(sb, KERN_ERR, "Could not load journal inode"); |
ac27a0ec DK |
3746 | iput(journal_inode); |
3747 | return NULL; | |
3748 | } | |
3749 | journal->j_private = sb; | |
617ba13b | 3750 | ext4_init_journal_params(sb, journal); |
ac27a0ec DK |
3751 | return journal; |
3752 | } | |
3753 | ||
617ba13b | 3754 | static journal_t *ext4_get_dev_journal(struct super_block *sb, |
ac27a0ec DK |
3755 | dev_t j_dev) |
3756 | { | |
2b2d6d01 | 3757 | struct buffer_head *bh; |
ac27a0ec | 3758 | journal_t *journal; |
617ba13b MC |
3759 | ext4_fsblk_t start; |
3760 | ext4_fsblk_t len; | |
ac27a0ec | 3761 | int hblock, blocksize; |
617ba13b | 3762 | ext4_fsblk_t sb_block; |
ac27a0ec | 3763 | unsigned long offset; |
2b2d6d01 | 3764 | struct ext4_super_block *es; |
ac27a0ec DK |
3765 | struct block_device *bdev; |
3766 | ||
0390131b FM |
3767 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
3768 | ||
b31e1552 | 3769 | bdev = ext4_blkdev_get(j_dev, sb); |
ac27a0ec DK |
3770 | if (bdev == NULL) |
3771 | return NULL; | |
3772 | ||
3773 | if (bd_claim(bdev, sb)) { | |
b31e1552 ES |
3774 | ext4_msg(sb, KERN_ERR, |
3775 | "failed to claim external journal device"); | |
9a1c3542 | 3776 | blkdev_put(bdev, FMODE_READ|FMODE_WRITE); |
ac27a0ec DK |
3777 | return NULL; |
3778 | } | |
3779 | ||
3780 | blocksize = sb->s_blocksize; | |
e1defc4f | 3781 | hblock = bdev_logical_block_size(bdev); |
ac27a0ec | 3782 | if (blocksize < hblock) { |
b31e1552 ES |
3783 | ext4_msg(sb, KERN_ERR, |
3784 | "blocksize too small for journal device"); | |
ac27a0ec DK |
3785 | goto out_bdev; |
3786 | } | |
3787 | ||
617ba13b MC |
3788 | sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; |
3789 | offset = EXT4_MIN_BLOCK_SIZE % blocksize; | |
ac27a0ec DK |
3790 | set_blocksize(bdev, blocksize); |
3791 | if (!(bh = __bread(bdev, sb_block, blocksize))) { | |
b31e1552 ES |
3792 | ext4_msg(sb, KERN_ERR, "couldn't read superblock of " |
3793 | "external journal"); | |
ac27a0ec DK |
3794 | goto out_bdev; |
3795 | } | |
3796 | ||
617ba13b MC |
3797 | es = (struct ext4_super_block *) (((char *)bh->b_data) + offset); |
3798 | if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || | |
ac27a0ec | 3799 | !(le32_to_cpu(es->s_feature_incompat) & |
617ba13b | 3800 | EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { |
b31e1552 ES |
3801 | ext4_msg(sb, KERN_ERR, "external journal has " |
3802 | "bad superblock"); | |
ac27a0ec DK |
3803 | brelse(bh); |
3804 | goto out_bdev; | |
3805 | } | |
3806 | ||
617ba13b | 3807 | if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { |
b31e1552 | 3808 | ext4_msg(sb, KERN_ERR, "journal UUID does not match"); |
ac27a0ec DK |
3809 | brelse(bh); |
3810 | goto out_bdev; | |
3811 | } | |
3812 | ||
bd81d8ee | 3813 | len = ext4_blocks_count(es); |
ac27a0ec DK |
3814 | start = sb_block + 1; |
3815 | brelse(bh); /* we're done with the superblock */ | |
3816 | ||
dab291af | 3817 | journal = jbd2_journal_init_dev(bdev, sb->s_bdev, |
ac27a0ec DK |
3818 | start, len, blocksize); |
3819 | if (!journal) { | |
b31e1552 | 3820 | ext4_msg(sb, KERN_ERR, "failed to create device journal"); |
ac27a0ec DK |
3821 | goto out_bdev; |
3822 | } | |
3823 | journal->j_private = sb; | |
3824 | ll_rw_block(READ, 1, &journal->j_sb_buffer); | |
3825 | wait_on_buffer(journal->j_sb_buffer); | |
3826 | if (!buffer_uptodate(journal->j_sb_buffer)) { | |
b31e1552 | 3827 | ext4_msg(sb, KERN_ERR, "I/O error on journal device"); |
ac27a0ec DK |
3828 | goto out_journal; |
3829 | } | |
3830 | if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { | |
b31e1552 ES |
3831 | ext4_msg(sb, KERN_ERR, "External journal has more than one " |
3832 | "user (unsupported) - %d", | |
ac27a0ec DK |
3833 | be32_to_cpu(journal->j_superblock->s_nr_users)); |
3834 | goto out_journal; | |
3835 | } | |
617ba13b MC |
3836 | EXT4_SB(sb)->journal_bdev = bdev; |
3837 | ext4_init_journal_params(sb, journal); | |
ac27a0ec | 3838 | return journal; |
0b8e58a1 | 3839 | |
ac27a0ec | 3840 | out_journal: |
dab291af | 3841 | jbd2_journal_destroy(journal); |
ac27a0ec | 3842 | out_bdev: |
617ba13b | 3843 | ext4_blkdev_put(bdev); |
ac27a0ec DK |
3844 | return NULL; |
3845 | } | |
3846 | ||
617ba13b MC |
3847 | static int ext4_load_journal(struct super_block *sb, |
3848 | struct ext4_super_block *es, | |
ac27a0ec DK |
3849 | unsigned long journal_devnum) |
3850 | { | |
3851 | journal_t *journal; | |
3852 | unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); | |
3853 | dev_t journal_dev; | |
3854 | int err = 0; | |
3855 | int really_read_only; | |
3856 | ||
0390131b FM |
3857 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
3858 | ||
ac27a0ec DK |
3859 | if (journal_devnum && |
3860 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { | |
b31e1552 ES |
3861 | ext4_msg(sb, KERN_INFO, "external journal device major/minor " |
3862 | "numbers have changed"); | |
ac27a0ec DK |
3863 | journal_dev = new_decode_dev(journal_devnum); |
3864 | } else | |
3865 | journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); | |
3866 | ||
3867 | really_read_only = bdev_read_only(sb->s_bdev); | |
3868 | ||
3869 | /* | |
3870 | * Are we loading a blank journal or performing recovery after a | |
3871 | * crash? For recovery, we need to check in advance whether we | |
3872 | * can get read-write access to the device. | |
3873 | */ | |
617ba13b | 3874 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { |
ac27a0ec | 3875 | if (sb->s_flags & MS_RDONLY) { |
b31e1552 ES |
3876 | ext4_msg(sb, KERN_INFO, "INFO: recovery " |
3877 | "required on readonly filesystem"); | |
ac27a0ec | 3878 | if (really_read_only) { |
b31e1552 ES |
3879 | ext4_msg(sb, KERN_ERR, "write access " |
3880 | "unavailable, cannot proceed"); | |
ac27a0ec DK |
3881 | return -EROFS; |
3882 | } | |
b31e1552 ES |
3883 | ext4_msg(sb, KERN_INFO, "write access will " |
3884 | "be enabled during recovery"); | |
ac27a0ec DK |
3885 | } |
3886 | } | |
3887 | ||
3888 | if (journal_inum && journal_dev) { | |
b31e1552 ES |
3889 | ext4_msg(sb, KERN_ERR, "filesystem has both journal " |
3890 | "and inode journals!"); | |
ac27a0ec DK |
3891 | return -EINVAL; |
3892 | } | |
3893 | ||
3894 | if (journal_inum) { | |
617ba13b | 3895 | if (!(journal = ext4_get_journal(sb, journal_inum))) |
ac27a0ec DK |
3896 | return -EINVAL; |
3897 | } else { | |
617ba13b | 3898 | if (!(journal = ext4_get_dev_journal(sb, journal_dev))) |
ac27a0ec DK |
3899 | return -EINVAL; |
3900 | } | |
3901 | ||
90576c0b | 3902 | if (!(journal->j_flags & JBD2_BARRIER)) |
b31e1552 | 3903 | ext4_msg(sb, KERN_INFO, "barriers disabled"); |
4776004f | 3904 | |
ac27a0ec | 3905 | if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) { |
dab291af | 3906 | err = jbd2_journal_update_format(journal); |
ac27a0ec | 3907 | if (err) { |
b31e1552 | 3908 | ext4_msg(sb, KERN_ERR, "error updating journal"); |
dab291af | 3909 | jbd2_journal_destroy(journal); |
ac27a0ec DK |
3910 | return err; |
3911 | } | |
3912 | } | |
3913 | ||
617ba13b | 3914 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) |
dab291af | 3915 | err = jbd2_journal_wipe(journal, !really_read_only); |
1c13d5c0 TT |
3916 | if (!err) { |
3917 | char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); | |
3918 | if (save) | |
3919 | memcpy(save, ((char *) es) + | |
3920 | EXT4_S_ERR_START, EXT4_S_ERR_LEN); | |
dab291af | 3921 | err = jbd2_journal_load(journal); |
1c13d5c0 TT |
3922 | if (save) |
3923 | memcpy(((char *) es) + EXT4_S_ERR_START, | |
3924 | save, EXT4_S_ERR_LEN); | |
3925 | kfree(save); | |
3926 | } | |
ac27a0ec DK |
3927 | |
3928 | if (err) { | |
b31e1552 | 3929 | ext4_msg(sb, KERN_ERR, "error loading journal"); |
dab291af | 3930 | jbd2_journal_destroy(journal); |
ac27a0ec DK |
3931 | return err; |
3932 | } | |
3933 | ||
617ba13b MC |
3934 | EXT4_SB(sb)->s_journal = journal; |
3935 | ext4_clear_journal_err(sb, es); | |
ac27a0ec | 3936 | |
c41303ce | 3937 | if (!really_read_only && journal_devnum && |
ac27a0ec DK |
3938 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { |
3939 | es->s_journal_dev = cpu_to_le32(journal_devnum); | |
ac27a0ec DK |
3940 | |
3941 | /* Make sure we flush the recovery flag to disk. */ | |
e2d67052 | 3942 | ext4_commit_super(sb, 1); |
ac27a0ec DK |
3943 | } |
3944 | ||
3945 | return 0; | |
3946 | } | |
3947 | ||
e2d67052 | 3948 | static int ext4_commit_super(struct super_block *sb, int sync) |
ac27a0ec | 3949 | { |
e2d67052 | 3950 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
617ba13b | 3951 | struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; |
c4be0c1d | 3952 | int error = 0; |
ac27a0ec DK |
3953 | |
3954 | if (!sbh) | |
c4be0c1d | 3955 | return error; |
914258bf TT |
3956 | if (buffer_write_io_error(sbh)) { |
3957 | /* | |
3958 | * Oh, dear. A previous attempt to write the | |
3959 | * superblock failed. This could happen because the | |
3960 | * USB device was yanked out. Or it could happen to | |
3961 | * be a transient write error and maybe the block will | |
3962 | * be remapped. Nothing we can do but to retry the | |
3963 | * write and hope for the best. | |
3964 | */ | |
b31e1552 ES |
3965 | ext4_msg(sb, KERN_ERR, "previous I/O error to " |
3966 | "superblock detected"); | |
914258bf TT |
3967 | clear_buffer_write_io_error(sbh); |
3968 | set_buffer_uptodate(sbh); | |
3969 | } | |
71290b36 TT |
3970 | /* |
3971 | * If the file system is mounted read-only, don't update the | |
3972 | * superblock write time. This avoids updating the superblock | |
3973 | * write time when we are mounting the root file system | |
3974 | * read/only but we need to replay the journal; at that point, | |
3975 | * for people who are east of GMT and who make their clock | |
3976 | * tick in localtime for Windows bug-for-bug compatibility, | |
3977 | * the clock is set in the future, and this will cause e2fsck | |
3978 | * to complain and force a full file system check. | |
3979 | */ | |
3980 | if (!(sb->s_flags & MS_RDONLY)) | |
3981 | es->s_wtime = cpu_to_le32(get_seconds()); | |
f613dfcb TT |
3982 | if (sb->s_bdev->bd_part) |
3983 | es->s_kbytes_written = | |
3984 | cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + | |
afc32f7e TT |
3985 | ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - |
3986 | EXT4_SB(sb)->s_sectors_written_start) >> 1)); | |
f613dfcb TT |
3987 | else |
3988 | es->s_kbytes_written = | |
3989 | cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); | |
ce7e010a TT |
3990 | ext4_free_blocks_count_set(es, percpu_counter_sum_positive( |
3991 | &EXT4_SB(sb)->s_freeblocks_counter)); | |
3992 | es->s_free_inodes_count = | |
3993 | cpu_to_le32(percpu_counter_sum_positive( | |
3994 | &EXT4_SB(sb)->s_freeinodes_counter)); | |
7234ab2a | 3995 | sb->s_dirt = 0; |
ac27a0ec DK |
3996 | BUFFER_TRACE(sbh, "marking dirty"); |
3997 | mark_buffer_dirty(sbh); | |
914258bf | 3998 | if (sync) { |
c4be0c1d TS |
3999 | error = sync_dirty_buffer(sbh); |
4000 | if (error) | |
4001 | return error; | |
4002 | ||
4003 | error = buffer_write_io_error(sbh); | |
4004 | if (error) { | |
b31e1552 ES |
4005 | ext4_msg(sb, KERN_ERR, "I/O error while writing " |
4006 | "superblock"); | |
914258bf TT |
4007 | clear_buffer_write_io_error(sbh); |
4008 | set_buffer_uptodate(sbh); | |
4009 | } | |
4010 | } | |
c4be0c1d | 4011 | return error; |
ac27a0ec DK |
4012 | } |
4013 | ||
ac27a0ec DK |
4014 | /* |
4015 | * Have we just finished recovery? If so, and if we are mounting (or | |
4016 | * remounting) the filesystem readonly, then we will end up with a | |
4017 | * consistent fs on disk. Record that fact. | |
4018 | */ | |
2b2d6d01 TT |
4019 | static void ext4_mark_recovery_complete(struct super_block *sb, |
4020 | struct ext4_super_block *es) | |
ac27a0ec | 4021 | { |
617ba13b | 4022 | journal_t *journal = EXT4_SB(sb)->s_journal; |
ac27a0ec | 4023 | |
0390131b FM |
4024 | if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { |
4025 | BUG_ON(journal != NULL); | |
4026 | return; | |
4027 | } | |
dab291af | 4028 | jbd2_journal_lock_updates(journal); |
7ffe1ea8 HK |
4029 | if (jbd2_journal_flush(journal) < 0) |
4030 | goto out; | |
4031 | ||
617ba13b | 4032 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) && |
ac27a0ec | 4033 | sb->s_flags & MS_RDONLY) { |
617ba13b | 4034 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); |
e2d67052 | 4035 | ext4_commit_super(sb, 1); |
ac27a0ec | 4036 | } |
7ffe1ea8 HK |
4037 | |
4038 | out: | |
dab291af | 4039 | jbd2_journal_unlock_updates(journal); |
ac27a0ec DK |
4040 | } |
4041 | ||
4042 | /* | |
4043 | * If we are mounting (or read-write remounting) a filesystem whose journal | |
4044 | * has recorded an error from a previous lifetime, move that error to the | |
4045 | * main filesystem now. | |
4046 | */ | |
2b2d6d01 TT |
4047 | static void ext4_clear_journal_err(struct super_block *sb, |
4048 | struct ext4_super_block *es) | |
ac27a0ec DK |
4049 | { |
4050 | journal_t *journal; | |
4051 | int j_errno; | |
4052 | const char *errstr; | |
4053 | ||
0390131b FM |
4054 | BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); |
4055 | ||
617ba13b | 4056 | journal = EXT4_SB(sb)->s_journal; |
ac27a0ec DK |
4057 | |
4058 | /* | |
4059 | * Now check for any error status which may have been recorded in the | |
617ba13b | 4060 | * journal by a prior ext4_error() or ext4_abort() |
ac27a0ec DK |
4061 | */ |
4062 | ||
dab291af | 4063 | j_errno = jbd2_journal_errno(journal); |
ac27a0ec DK |
4064 | if (j_errno) { |
4065 | char nbuf[16]; | |
4066 | ||
617ba13b | 4067 | errstr = ext4_decode_error(sb, j_errno, nbuf); |
12062ddd | 4068 | ext4_warning(sb, "Filesystem error recorded " |
ac27a0ec | 4069 | "from previous mount: %s", errstr); |
12062ddd | 4070 | ext4_warning(sb, "Marking fs in need of filesystem check."); |
ac27a0ec | 4071 | |
617ba13b MC |
4072 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; |
4073 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
e2d67052 | 4074 | ext4_commit_super(sb, 1); |
ac27a0ec | 4075 | |
dab291af | 4076 | jbd2_journal_clear_err(journal); |
ac27a0ec DK |
4077 | } |
4078 | } | |
4079 | ||
4080 | /* | |
4081 | * Force the running and committing transactions to commit, | |
4082 | * and wait on the commit. | |
4083 | */ | |
617ba13b | 4084 | int ext4_force_commit(struct super_block *sb) |
ac27a0ec DK |
4085 | { |
4086 | journal_t *journal; | |
0390131b | 4087 | int ret = 0; |
ac27a0ec DK |
4088 | |
4089 | if (sb->s_flags & MS_RDONLY) | |
4090 | return 0; | |
4091 | ||
617ba13b | 4092 | journal = EXT4_SB(sb)->s_journal; |
6b0310fb | 4093 | if (journal) { |
437f88cc | 4094 | vfs_check_frozen(sb, SB_FREEZE_TRANS); |
0390131b | 4095 | ret = ext4_journal_force_commit(journal); |
6b0310fb | 4096 | } |
0390131b | 4097 | |
ac27a0ec DK |
4098 | return ret; |
4099 | } | |
4100 | ||
2b2d6d01 | 4101 | static void ext4_write_super(struct super_block *sb) |
ac27a0ec | 4102 | { |
ebc1ac16 | 4103 | lock_super(sb); |
9ca92389 | 4104 | ext4_commit_super(sb, 1); |
ebc1ac16 | 4105 | unlock_super(sb); |
ac27a0ec DK |
4106 | } |
4107 | ||
617ba13b | 4108 | static int ext4_sync_fs(struct super_block *sb, int wait) |
ac27a0ec | 4109 | { |
14ce0cb4 | 4110 | int ret = 0; |
9eddacf9 | 4111 | tid_t target; |
8d5d02e6 | 4112 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 4113 | |
9bffad1e | 4114 | trace_ext4_sync_fs(sb, wait); |
8d5d02e6 MC |
4115 | flush_workqueue(sbi->dio_unwritten_wq); |
4116 | if (jbd2_journal_start_commit(sbi->s_journal, &target)) { | |
9ca92389 | 4117 | if (wait) |
8d5d02e6 | 4118 | jbd2_log_wait_commit(sbi->s_journal, target); |
0390131b | 4119 | } |
14ce0cb4 | 4120 | return ret; |
ac27a0ec DK |
4121 | } |
4122 | ||
4123 | /* | |
4124 | * LVM calls this function before a (read-only) snapshot is created. This | |
4125 | * gives us a chance to flush the journal completely and mark the fs clean. | |
4126 | */ | |
c4be0c1d | 4127 | static int ext4_freeze(struct super_block *sb) |
ac27a0ec | 4128 | { |
c4be0c1d TS |
4129 | int error = 0; |
4130 | journal_t *journal; | |
ac27a0ec | 4131 | |
9ca92389 TT |
4132 | if (sb->s_flags & MS_RDONLY) |
4133 | return 0; | |
ac27a0ec | 4134 | |
9ca92389 | 4135 | journal = EXT4_SB(sb)->s_journal; |
7ffe1ea8 | 4136 | |
9ca92389 TT |
4137 | /* Now we set up the journal barrier. */ |
4138 | jbd2_journal_lock_updates(journal); | |
ac27a0ec | 4139 | |
9ca92389 TT |
4140 | /* |
4141 | * Don't clear the needs_recovery flag if we failed to flush | |
4142 | * the journal. | |
4143 | */ | |
4144 | error = jbd2_journal_flush(journal); | |
6b0310fb ES |
4145 | if (error < 0) |
4146 | goto out; | |
9ca92389 TT |
4147 | |
4148 | /* Journal blocked and flushed, clear needs_recovery flag. */ | |
4149 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
4150 | error = ext4_commit_super(sb, 1); | |
6b0310fb ES |
4151 | out: |
4152 | /* we rely on s_frozen to stop further updates */ | |
4153 | jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); | |
4154 | return error; | |
ac27a0ec DK |
4155 | } |
4156 | ||
4157 | /* | |
4158 | * Called by LVM after the snapshot is done. We need to reset the RECOVER | |
4159 | * flag here, even though the filesystem is not technically dirty yet. | |
4160 | */ | |
c4be0c1d | 4161 | static int ext4_unfreeze(struct super_block *sb) |
ac27a0ec | 4162 | { |
9ca92389 TT |
4163 | if (sb->s_flags & MS_RDONLY) |
4164 | return 0; | |
4165 | ||
4166 | lock_super(sb); | |
4167 | /* Reset the needs_recovery flag before the fs is unlocked. */ | |
4168 | EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
4169 | ext4_commit_super(sb, 1); | |
4170 | unlock_super(sb); | |
c4be0c1d | 4171 | return 0; |
ac27a0ec DK |
4172 | } |
4173 | ||
673c6100 TT |
4174 | /* |
4175 | * Structure to save mount options for ext4_remount's benefit | |
4176 | */ | |
4177 | struct ext4_mount_options { | |
4178 | unsigned long s_mount_opt; | |
a2595b8a | 4179 | unsigned long s_mount_opt2; |
673c6100 TT |
4180 | uid_t s_resuid; |
4181 | gid_t s_resgid; | |
4182 | unsigned long s_commit_interval; | |
4183 | u32 s_min_batch_time, s_max_batch_time; | |
4184 | #ifdef CONFIG_QUOTA | |
4185 | int s_jquota_fmt; | |
4186 | char *s_qf_names[MAXQUOTAS]; | |
4187 | #endif | |
4188 | }; | |
4189 | ||
2b2d6d01 | 4190 | static int ext4_remount(struct super_block *sb, int *flags, char *data) |
ac27a0ec | 4191 | { |
2b2d6d01 | 4192 | struct ext4_super_block *es; |
617ba13b MC |
4193 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
4194 | ext4_fsblk_t n_blocks_count = 0; | |
ac27a0ec | 4195 | unsigned long old_sb_flags; |
617ba13b | 4196 | struct ext4_mount_options old_opts; |
c79d967d | 4197 | int enable_quota = 0; |
8a266467 | 4198 | ext4_group_t g; |
b3881f74 | 4199 | unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; |
ac27a0ec DK |
4200 | int err; |
4201 | #ifdef CONFIG_QUOTA | |
4202 | int i; | |
4203 | #endif | |
d4c402d9 | 4204 | char *orig_data = kstrdup(data, GFP_KERNEL); |
ac27a0ec DK |
4205 | |
4206 | /* Store the original options */ | |
bbd6851a | 4207 | lock_super(sb); |
ac27a0ec DK |
4208 | old_sb_flags = sb->s_flags; |
4209 | old_opts.s_mount_opt = sbi->s_mount_opt; | |
a2595b8a | 4210 | old_opts.s_mount_opt2 = sbi->s_mount_opt2; |
ac27a0ec DK |
4211 | old_opts.s_resuid = sbi->s_resuid; |
4212 | old_opts.s_resgid = sbi->s_resgid; | |
4213 | old_opts.s_commit_interval = sbi->s_commit_interval; | |
30773840 TT |
4214 | old_opts.s_min_batch_time = sbi->s_min_batch_time; |
4215 | old_opts.s_max_batch_time = sbi->s_max_batch_time; | |
ac27a0ec DK |
4216 | #ifdef CONFIG_QUOTA |
4217 | old_opts.s_jquota_fmt = sbi->s_jquota_fmt; | |
4218 | for (i = 0; i < MAXQUOTAS; i++) | |
4219 | old_opts.s_qf_names[i] = sbi->s_qf_names[i]; | |
4220 | #endif | |
b3881f74 TT |
4221 | if (sbi->s_journal && sbi->s_journal->j_task->io_context) |
4222 | journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; | |
ac27a0ec DK |
4223 | |
4224 | /* | |
4225 | * Allow the "check" option to be passed as a remount option. | |
4226 | */ | |
b3881f74 TT |
4227 | if (!parse_options(data, sb, NULL, &journal_ioprio, |
4228 | &n_blocks_count, 1)) { | |
ac27a0ec DK |
4229 | err = -EINVAL; |
4230 | goto restore_opts; | |
4231 | } | |
4232 | ||
4ab2f15b | 4233 | if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) |
c67d859e | 4234 | ext4_abort(sb, "Abort forced by user"); |
ac27a0ec DK |
4235 | |
4236 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
482a7425 | 4237 | (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); |
ac27a0ec DK |
4238 | |
4239 | es = sbi->s_es; | |
4240 | ||
b3881f74 | 4241 | if (sbi->s_journal) { |
0390131b | 4242 | ext4_init_journal_params(sb, sbi->s_journal); |
b3881f74 TT |
4243 | set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); |
4244 | } | |
ac27a0ec DK |
4245 | |
4246 | if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) || | |
bd81d8ee | 4247 | n_blocks_count > ext4_blocks_count(es)) { |
4ab2f15b | 4248 | if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { |
ac27a0ec DK |
4249 | err = -EROFS; |
4250 | goto restore_opts; | |
4251 | } | |
4252 | ||
4253 | if (*flags & MS_RDONLY) { | |
0f0dd62f CH |
4254 | err = dquot_suspend(sb, -1); |
4255 | if (err < 0) | |
c79d967d | 4256 | goto restore_opts; |
c79d967d | 4257 | |
ac27a0ec DK |
4258 | /* |
4259 | * First of all, the unconditional stuff we have to do | |
4260 | * to disable replay of the journal when we next remount | |
4261 | */ | |
4262 | sb->s_flags |= MS_RDONLY; | |
4263 | ||
4264 | /* | |
4265 | * OK, test if we are remounting a valid rw partition | |
4266 | * readonly, and if so set the rdonly flag and then | |
4267 | * mark the partition as valid again. | |
4268 | */ | |
617ba13b MC |
4269 | if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && |
4270 | (sbi->s_mount_state & EXT4_VALID_FS)) | |
ac27a0ec DK |
4271 | es->s_state = cpu_to_le16(sbi->s_mount_state); |
4272 | ||
a63c9eb2 | 4273 | if (sbi->s_journal) |
0390131b | 4274 | ext4_mark_recovery_complete(sb, es); |
ac27a0ec | 4275 | } else { |
a13fb1a4 ES |
4276 | /* Make sure we can mount this feature set readwrite */ |
4277 | if (!ext4_feature_set_ok(sb, 0)) { | |
ac27a0ec DK |
4278 | err = -EROFS; |
4279 | goto restore_opts; | |
4280 | } | |
8a266467 TT |
4281 | /* |
4282 | * Make sure the group descriptor checksums | |
0b8e58a1 | 4283 | * are sane. If they aren't, refuse to remount r/w. |
8a266467 TT |
4284 | */ |
4285 | for (g = 0; g < sbi->s_groups_count; g++) { | |
4286 | struct ext4_group_desc *gdp = | |
4287 | ext4_get_group_desc(sb, g, NULL); | |
4288 | ||
4289 | if (!ext4_group_desc_csum_verify(sbi, g, gdp)) { | |
b31e1552 ES |
4290 | ext4_msg(sb, KERN_ERR, |
4291 | "ext4_remount: Checksum for group %u failed (%u!=%u)", | |
8a266467 TT |
4292 | g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)), |
4293 | le16_to_cpu(gdp->bg_checksum)); | |
4294 | err = -EINVAL; | |
4295 | goto restore_opts; | |
4296 | } | |
4297 | } | |
4298 | ||
ead6596b ES |
4299 | /* |
4300 | * If we have an unprocessed orphan list hanging | |
4301 | * around from a previously readonly bdev mount, | |
4302 | * require a full umount/remount for now. | |
4303 | */ | |
4304 | if (es->s_last_orphan) { | |
b31e1552 | 4305 | ext4_msg(sb, KERN_WARNING, "Couldn't " |
ead6596b ES |
4306 | "remount RDWR because of unprocessed " |
4307 | "orphan inode list. Please " | |
b31e1552 | 4308 | "umount/remount instead"); |
ead6596b ES |
4309 | err = -EINVAL; |
4310 | goto restore_opts; | |
4311 | } | |
4312 | ||
ac27a0ec DK |
4313 | /* |
4314 | * Mounting a RDONLY partition read-write, so reread | |
4315 | * and store the current valid flag. (It may have | |
4316 | * been changed by e2fsck since we originally mounted | |
4317 | * the partition.) | |
4318 | */ | |
0390131b FM |
4319 | if (sbi->s_journal) |
4320 | ext4_clear_journal_err(sb, es); | |
ac27a0ec | 4321 | sbi->s_mount_state = le16_to_cpu(es->s_state); |
617ba13b | 4322 | if ((err = ext4_group_extend(sb, es, n_blocks_count))) |
ac27a0ec | 4323 | goto restore_opts; |
2b2d6d01 | 4324 | if (!ext4_setup_super(sb, es, 0)) |
ac27a0ec | 4325 | sb->s_flags &= ~MS_RDONLY; |
c79d967d | 4326 | enable_quota = 1; |
ac27a0ec DK |
4327 | } |
4328 | } | |
bfff6873 LC |
4329 | |
4330 | /* | |
4331 | * Reinitialize lazy itable initialization thread based on | |
4332 | * current settings | |
4333 | */ | |
4334 | if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE)) | |
4335 | ext4_unregister_li_request(sb); | |
4336 | else { | |
4337 | ext4_group_t first_not_zeroed; | |
4338 | first_not_zeroed = ext4_has_uninit_itable(sb); | |
4339 | ext4_register_li_request(sb, first_not_zeroed); | |
4340 | } | |
4341 | ||
6fd058f7 | 4342 | ext4_setup_system_zone(sb); |
0390131b | 4343 | if (sbi->s_journal == NULL) |
e2d67052 | 4344 | ext4_commit_super(sb, 1); |
0390131b | 4345 | |
ac27a0ec DK |
4346 | #ifdef CONFIG_QUOTA |
4347 | /* Release old quota file names */ | |
4348 | for (i = 0; i < MAXQUOTAS; i++) | |
4349 | if (old_opts.s_qf_names[i] && | |
4350 | old_opts.s_qf_names[i] != sbi->s_qf_names[i]) | |
4351 | kfree(old_opts.s_qf_names[i]); | |
4352 | #endif | |
bbd6851a | 4353 | unlock_super(sb); |
c79d967d | 4354 | if (enable_quota) |
0f0dd62f | 4355 | dquot_resume(sb, -1); |
d4c402d9 CW |
4356 | |
4357 | ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); | |
4358 | kfree(orig_data); | |
ac27a0ec | 4359 | return 0; |
0b8e58a1 | 4360 | |
ac27a0ec DK |
4361 | restore_opts: |
4362 | sb->s_flags = old_sb_flags; | |
4363 | sbi->s_mount_opt = old_opts.s_mount_opt; | |
a2595b8a | 4364 | sbi->s_mount_opt2 = old_opts.s_mount_opt2; |
ac27a0ec DK |
4365 | sbi->s_resuid = old_opts.s_resuid; |
4366 | sbi->s_resgid = old_opts.s_resgid; | |
4367 | sbi->s_commit_interval = old_opts.s_commit_interval; | |
30773840 TT |
4368 | sbi->s_min_batch_time = old_opts.s_min_batch_time; |
4369 | sbi->s_max_batch_time = old_opts.s_max_batch_time; | |
ac27a0ec DK |
4370 | #ifdef CONFIG_QUOTA |
4371 | sbi->s_jquota_fmt = old_opts.s_jquota_fmt; | |
4372 | for (i = 0; i < MAXQUOTAS; i++) { | |
4373 | if (sbi->s_qf_names[i] && | |
4374 | old_opts.s_qf_names[i] != sbi->s_qf_names[i]) | |
4375 | kfree(sbi->s_qf_names[i]); | |
4376 | sbi->s_qf_names[i] = old_opts.s_qf_names[i]; | |
4377 | } | |
4378 | #endif | |
bbd6851a | 4379 | unlock_super(sb); |
d4c402d9 | 4380 | kfree(orig_data); |
ac27a0ec DK |
4381 | return err; |
4382 | } | |
4383 | ||
2b2d6d01 | 4384 | static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) |
ac27a0ec DK |
4385 | { |
4386 | struct super_block *sb = dentry->d_sb; | |
617ba13b MC |
4387 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
4388 | struct ext4_super_block *es = sbi->s_es; | |
960cc398 | 4389 | u64 fsid; |
ac27a0ec | 4390 | |
5e70030d BP |
4391 | if (test_opt(sb, MINIX_DF)) { |
4392 | sbi->s_overhead_last = 0; | |
6bc9feff | 4393 | } else if (sbi->s_blocks_last != ext4_blocks_count(es)) { |
8df9675f | 4394 | ext4_group_t i, ngroups = ext4_get_groups_count(sb); |
5e70030d | 4395 | ext4_fsblk_t overhead = 0; |
ac27a0ec DK |
4396 | |
4397 | /* | |
5e70030d BP |
4398 | * Compute the overhead (FS structures). This is constant |
4399 | * for a given filesystem unless the number of block groups | |
4400 | * changes so we cache the previous value until it does. | |
ac27a0ec DK |
4401 | */ |
4402 | ||
4403 | /* | |
4404 | * All of the blocks before first_data_block are | |
4405 | * overhead | |
4406 | */ | |
4407 | overhead = le32_to_cpu(es->s_first_data_block); | |
4408 | ||
4409 | /* | |
4410 | * Add the overhead attributed to the superblock and | |
4411 | * block group descriptors. If the sparse superblocks | |
4412 | * feature is turned on, then not all groups have this. | |
4413 | */ | |
4414 | for (i = 0; i < ngroups; i++) { | |
617ba13b MC |
4415 | overhead += ext4_bg_has_super(sb, i) + |
4416 | ext4_bg_num_gdb(sb, i); | |
ac27a0ec DK |
4417 | cond_resched(); |
4418 | } | |
4419 | ||
4420 | /* | |
4421 | * Every block group has an inode bitmap, a block | |
4422 | * bitmap, and an inode table. | |
4423 | */ | |
5e70030d BP |
4424 | overhead += ngroups * (2 + sbi->s_itb_per_group); |
4425 | sbi->s_overhead_last = overhead; | |
4426 | smp_wmb(); | |
6bc9feff | 4427 | sbi->s_blocks_last = ext4_blocks_count(es); |
ac27a0ec DK |
4428 | } |
4429 | ||
617ba13b | 4430 | buf->f_type = EXT4_SUPER_MAGIC; |
ac27a0ec | 4431 | buf->f_bsize = sb->s_blocksize; |
5e70030d | 4432 | buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last; |
6bc6e63f AK |
4433 | buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) - |
4434 | percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter); | |
bd81d8ee LV |
4435 | buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es); |
4436 | if (buf->f_bfree < ext4_r_blocks_count(es)) | |
ac27a0ec DK |
4437 | buf->f_bavail = 0; |
4438 | buf->f_files = le32_to_cpu(es->s_inodes_count); | |
52d9f3b4 | 4439 | buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); |
617ba13b | 4440 | buf->f_namelen = EXT4_NAME_LEN; |
960cc398 PE |
4441 | fsid = le64_to_cpup((void *)es->s_uuid) ^ |
4442 | le64_to_cpup((void *)es->s_uuid + sizeof(u64)); | |
4443 | buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; | |
4444 | buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; | |
0b8e58a1 | 4445 | |
ac27a0ec DK |
4446 | return 0; |
4447 | } | |
4448 | ||
0b8e58a1 AD |
4449 | /* Helper function for writing quotas on sync - we need to start transaction |
4450 | * before quota file is locked for write. Otherwise the are possible deadlocks: | |
ac27a0ec | 4451 | * Process 1 Process 2 |
617ba13b | 4452 | * ext4_create() quota_sync() |
a269eb18 | 4453 | * jbd2_journal_start() write_dquot() |
871a2931 | 4454 | * dquot_initialize() down(dqio_mutex) |
dab291af | 4455 | * down(dqio_mutex) jbd2_journal_start() |
ac27a0ec DK |
4456 | * |
4457 | */ | |
4458 | ||
4459 | #ifdef CONFIG_QUOTA | |
4460 | ||
4461 | static inline struct inode *dquot_to_inode(struct dquot *dquot) | |
4462 | { | |
4463 | return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type]; | |
4464 | } | |
4465 | ||
617ba13b | 4466 | static int ext4_write_dquot(struct dquot *dquot) |
ac27a0ec DK |
4467 | { |
4468 | int ret, err; | |
4469 | handle_t *handle; | |
4470 | struct inode *inode; | |
4471 | ||
4472 | inode = dquot_to_inode(dquot); | |
617ba13b | 4473 | handle = ext4_journal_start(inode, |
0b8e58a1 | 4474 | EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); |
ac27a0ec DK |
4475 | if (IS_ERR(handle)) |
4476 | return PTR_ERR(handle); | |
4477 | ret = dquot_commit(dquot); | |
617ba13b | 4478 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
4479 | if (!ret) |
4480 | ret = err; | |
4481 | return ret; | |
4482 | } | |
4483 | ||
617ba13b | 4484 | static int ext4_acquire_dquot(struct dquot *dquot) |
ac27a0ec DK |
4485 | { |
4486 | int ret, err; | |
4487 | handle_t *handle; | |
4488 | ||
617ba13b | 4489 | handle = ext4_journal_start(dquot_to_inode(dquot), |
0b8e58a1 | 4490 | EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); |
ac27a0ec DK |
4491 | if (IS_ERR(handle)) |
4492 | return PTR_ERR(handle); | |
4493 | ret = dquot_acquire(dquot); | |
617ba13b | 4494 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
4495 | if (!ret) |
4496 | ret = err; | |
4497 | return ret; | |
4498 | } | |
4499 | ||
617ba13b | 4500 | static int ext4_release_dquot(struct dquot *dquot) |
ac27a0ec DK |
4501 | { |
4502 | int ret, err; | |
4503 | handle_t *handle; | |
4504 | ||
617ba13b | 4505 | handle = ext4_journal_start(dquot_to_inode(dquot), |
0b8e58a1 | 4506 | EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); |
9c3013e9 JK |
4507 | if (IS_ERR(handle)) { |
4508 | /* Release dquot anyway to avoid endless cycle in dqput() */ | |
4509 | dquot_release(dquot); | |
ac27a0ec | 4510 | return PTR_ERR(handle); |
9c3013e9 | 4511 | } |
ac27a0ec | 4512 | ret = dquot_release(dquot); |
617ba13b | 4513 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
4514 | if (!ret) |
4515 | ret = err; | |
4516 | return ret; | |
4517 | } | |
4518 | ||
617ba13b | 4519 | static int ext4_mark_dquot_dirty(struct dquot *dquot) |
ac27a0ec | 4520 | { |
2c8be6b2 | 4521 | /* Are we journaling quotas? */ |
617ba13b MC |
4522 | if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] || |
4523 | EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) { | |
ac27a0ec | 4524 | dquot_mark_dquot_dirty(dquot); |
617ba13b | 4525 | return ext4_write_dquot(dquot); |
ac27a0ec DK |
4526 | } else { |
4527 | return dquot_mark_dquot_dirty(dquot); | |
4528 | } | |
4529 | } | |
4530 | ||
617ba13b | 4531 | static int ext4_write_info(struct super_block *sb, int type) |
ac27a0ec DK |
4532 | { |
4533 | int ret, err; | |
4534 | handle_t *handle; | |
4535 | ||
4536 | /* Data block + inode block */ | |
617ba13b | 4537 | handle = ext4_journal_start(sb->s_root->d_inode, 2); |
ac27a0ec DK |
4538 | if (IS_ERR(handle)) |
4539 | return PTR_ERR(handle); | |
4540 | ret = dquot_commit_info(sb, type); | |
617ba13b | 4541 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
4542 | if (!ret) |
4543 | ret = err; | |
4544 | return ret; | |
4545 | } | |
4546 | ||
4547 | /* | |
4548 | * Turn on quotas during mount time - we need to find | |
4549 | * the quota file and such... | |
4550 | */ | |
617ba13b | 4551 | static int ext4_quota_on_mount(struct super_block *sb, int type) |
ac27a0ec | 4552 | { |
287a8095 CH |
4553 | return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type], |
4554 | EXT4_SB(sb)->s_jquota_fmt, type); | |
ac27a0ec DK |
4555 | } |
4556 | ||
4557 | /* | |
4558 | * Standard function to be called on quota_on | |
4559 | */ | |
617ba13b | 4560 | static int ext4_quota_on(struct super_block *sb, int type, int format_id, |
307ae18a | 4561 | char *name) |
ac27a0ec DK |
4562 | { |
4563 | int err; | |
8264613d | 4564 | struct path path; |
ac27a0ec DK |
4565 | |
4566 | if (!test_opt(sb, QUOTA)) | |
4567 | return -EINVAL; | |
0623543b | 4568 | |
8264613d | 4569 | err = kern_path(name, LOOKUP_FOLLOW, &path); |
ac27a0ec DK |
4570 | if (err) |
4571 | return err; | |
0623543b | 4572 | |
ac27a0ec | 4573 | /* Quotafile not on the same filesystem? */ |
8264613d AV |
4574 | if (path.mnt->mnt_sb != sb) { |
4575 | path_put(&path); | |
ac27a0ec DK |
4576 | return -EXDEV; |
4577 | } | |
0623543b JK |
4578 | /* Journaling quota? */ |
4579 | if (EXT4_SB(sb)->s_qf_names[type]) { | |
2b2d6d01 | 4580 | /* Quotafile not in fs root? */ |
8264613d | 4581 | if (path.dentry->d_parent != sb->s_root) |
b31e1552 ES |
4582 | ext4_msg(sb, KERN_WARNING, |
4583 | "Quota file not on filesystem root. " | |
4584 | "Journaled quota will not work"); | |
2b2d6d01 | 4585 | } |
0623543b JK |
4586 | |
4587 | /* | |
4588 | * When we journal data on quota file, we have to flush journal to see | |
4589 | * all updates to the file when we bypass pagecache... | |
4590 | */ | |
0390131b FM |
4591 | if (EXT4_SB(sb)->s_journal && |
4592 | ext4_should_journal_data(path.dentry->d_inode)) { | |
0623543b JK |
4593 | /* |
4594 | * We don't need to lock updates but journal_flush() could | |
4595 | * otherwise be livelocked... | |
4596 | */ | |
4597 | jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); | |
7ffe1ea8 | 4598 | err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); |
0623543b | 4599 | jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); |
7ffe1ea8 | 4600 | if (err) { |
8264613d | 4601 | path_put(&path); |
7ffe1ea8 HK |
4602 | return err; |
4603 | } | |
0623543b JK |
4604 | } |
4605 | ||
287a8095 | 4606 | err = dquot_quota_on_path(sb, type, format_id, &path); |
8264613d | 4607 | path_put(&path); |
77e69dac | 4608 | return err; |
ac27a0ec DK |
4609 | } |
4610 | ||
ca0e05e4 DM |
4611 | static int ext4_quota_off(struct super_block *sb, int type) |
4612 | { | |
87009d86 DM |
4613 | /* Force all delayed allocation blocks to be allocated. |
4614 | * Caller already holds s_umount sem */ | |
4615 | if (test_opt(sb, DELALLOC)) | |
ca0e05e4 | 4616 | sync_filesystem(sb); |
ca0e05e4 DM |
4617 | |
4618 | return dquot_quota_off(sb, type); | |
4619 | } | |
4620 | ||
ac27a0ec DK |
4621 | /* Read data from quotafile - avoid pagecache and such because we cannot afford |
4622 | * acquiring the locks... As quota files are never truncated and quota code | |
4623 | * itself serializes the operations (and noone else should touch the files) | |
4624 | * we don't have to be afraid of races */ | |
617ba13b | 4625 | static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, |
ac27a0ec DK |
4626 | size_t len, loff_t off) |
4627 | { | |
4628 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
725d26d3 | 4629 | ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); |
ac27a0ec DK |
4630 | int err = 0; |
4631 | int offset = off & (sb->s_blocksize - 1); | |
4632 | int tocopy; | |
4633 | size_t toread; | |
4634 | struct buffer_head *bh; | |
4635 | loff_t i_size = i_size_read(inode); | |
4636 | ||
4637 | if (off > i_size) | |
4638 | return 0; | |
4639 | if (off+len > i_size) | |
4640 | len = i_size-off; | |
4641 | toread = len; | |
4642 | while (toread > 0) { | |
4643 | tocopy = sb->s_blocksize - offset < toread ? | |
4644 | sb->s_blocksize - offset : toread; | |
617ba13b | 4645 | bh = ext4_bread(NULL, inode, blk, 0, &err); |
ac27a0ec DK |
4646 | if (err) |
4647 | return err; | |
4648 | if (!bh) /* A hole? */ | |
4649 | memset(data, 0, tocopy); | |
4650 | else | |
4651 | memcpy(data, bh->b_data+offset, tocopy); | |
4652 | brelse(bh); | |
4653 | offset = 0; | |
4654 | toread -= tocopy; | |
4655 | data += tocopy; | |
4656 | blk++; | |
4657 | } | |
4658 | return len; | |
4659 | } | |
4660 | ||
4661 | /* Write to quotafile (we know the transaction is already started and has | |
4662 | * enough credits) */ | |
617ba13b | 4663 | static ssize_t ext4_quota_write(struct super_block *sb, int type, |
ac27a0ec DK |
4664 | const char *data, size_t len, loff_t off) |
4665 | { | |
4666 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
725d26d3 | 4667 | ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); |
ac27a0ec DK |
4668 | int err = 0; |
4669 | int offset = off & (sb->s_blocksize - 1); | |
ac27a0ec DK |
4670 | struct buffer_head *bh; |
4671 | handle_t *handle = journal_current_handle(); | |
4672 | ||
0390131b | 4673 | if (EXT4_SB(sb)->s_journal && !handle) { |
b31e1552 ES |
4674 | ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" |
4675 | " cancelled because transaction is not started", | |
9c3013e9 JK |
4676 | (unsigned long long)off, (unsigned long long)len); |
4677 | return -EIO; | |
4678 | } | |
67eeb568 DM |
4679 | /* |
4680 | * Since we account only one data block in transaction credits, | |
4681 | * then it is impossible to cross a block boundary. | |
4682 | */ | |
4683 | if (sb->s_blocksize - offset < len) { | |
4684 | ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" | |
4685 | " cancelled because not block aligned", | |
4686 | (unsigned long long)off, (unsigned long long)len); | |
4687 | return -EIO; | |
4688 | } | |
4689 | ||
ac27a0ec | 4690 | mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA); |
67eeb568 DM |
4691 | bh = ext4_bread(handle, inode, blk, 1, &err); |
4692 | if (!bh) | |
4693 | goto out; | |
62d2b5f2 JK |
4694 | err = ext4_journal_get_write_access(handle, bh); |
4695 | if (err) { | |
4696 | brelse(bh); | |
4697 | goto out; | |
ac27a0ec | 4698 | } |
67eeb568 DM |
4699 | lock_buffer(bh); |
4700 | memcpy(bh->b_data+offset, data, len); | |
4701 | flush_dcache_page(bh->b_page); | |
4702 | unlock_buffer(bh); | |
62d2b5f2 | 4703 | err = ext4_handle_dirty_metadata(handle, NULL, bh); |
67eeb568 | 4704 | brelse(bh); |
ac27a0ec | 4705 | out: |
67eeb568 | 4706 | if (err) { |
4d04e4fb | 4707 | mutex_unlock(&inode->i_mutex); |
ac27a0ec | 4708 | return err; |
4d04e4fb | 4709 | } |
67eeb568 DM |
4710 | if (inode->i_size < off + len) { |
4711 | i_size_write(inode, off + len); | |
617ba13b | 4712 | EXT4_I(inode)->i_disksize = inode->i_size; |
ac27a0ec | 4713 | } |
ac27a0ec | 4714 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; |
617ba13b | 4715 | ext4_mark_inode_dirty(handle, inode); |
ac27a0ec | 4716 | mutex_unlock(&inode->i_mutex); |
67eeb568 | 4717 | return len; |
ac27a0ec DK |
4718 | } |
4719 | ||
4720 | #endif | |
4721 | ||
152a0836 AV |
4722 | static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, |
4723 | const char *dev_name, void *data) | |
ac27a0ec | 4724 | { |
152a0836 | 4725 | return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); |
ac27a0ec DK |
4726 | } |
4727 | ||
37f328eb | 4728 | #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) |
24b58424 TT |
4729 | static struct file_system_type ext2_fs_type = { |
4730 | .owner = THIS_MODULE, | |
4731 | .name = "ext2", | |
152a0836 | 4732 | .mount = ext4_mount, |
24b58424 TT |
4733 | .kill_sb = kill_block_super, |
4734 | .fs_flags = FS_REQUIRES_DEV, | |
4735 | }; | |
4736 | ||
4737 | static inline void register_as_ext2(void) | |
4738 | { | |
4739 | int err = register_filesystem(&ext2_fs_type); | |
4740 | if (err) | |
4741 | printk(KERN_WARNING | |
4742 | "EXT4-fs: Unable to register as ext2 (%d)\n", err); | |
4743 | } | |
4744 | ||
4745 | static inline void unregister_as_ext2(void) | |
4746 | { | |
4747 | unregister_filesystem(&ext2_fs_type); | |
4748 | } | |
51b7e3c9 | 4749 | MODULE_ALIAS("ext2"); |
24b58424 TT |
4750 | #else |
4751 | static inline void register_as_ext2(void) { } | |
4752 | static inline void unregister_as_ext2(void) { } | |
4753 | #endif | |
4754 | ||
37f328eb | 4755 | #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) |
24b58424 TT |
4756 | static inline void register_as_ext3(void) |
4757 | { | |
4758 | int err = register_filesystem(&ext3_fs_type); | |
4759 | if (err) | |
4760 | printk(KERN_WARNING | |
4761 | "EXT4-fs: Unable to register as ext3 (%d)\n", err); | |
4762 | } | |
4763 | ||
4764 | static inline void unregister_as_ext3(void) | |
4765 | { | |
4766 | unregister_filesystem(&ext3_fs_type); | |
4767 | } | |
51b7e3c9 | 4768 | MODULE_ALIAS("ext3"); |
24b58424 TT |
4769 | #else |
4770 | static inline void register_as_ext3(void) { } | |
4771 | static inline void unregister_as_ext3(void) { } | |
4772 | #endif | |
4773 | ||
03010a33 TT |
4774 | static struct file_system_type ext4_fs_type = { |
4775 | .owner = THIS_MODULE, | |
4776 | .name = "ext4", | |
152a0836 | 4777 | .mount = ext4_mount, |
03010a33 TT |
4778 | .kill_sb = kill_block_super, |
4779 | .fs_flags = FS_REQUIRES_DEV, | |
4780 | }; | |
4781 | ||
857ac889 LC |
4782 | int __init ext4_init_feat_adverts(void) |
4783 | { | |
4784 | struct ext4_features *ef; | |
4785 | int ret = -ENOMEM; | |
4786 | ||
4787 | ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL); | |
4788 | if (!ef) | |
4789 | goto out; | |
4790 | ||
4791 | ef->f_kobj.kset = ext4_kset; | |
4792 | init_completion(&ef->f_kobj_unregister); | |
4793 | ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL, | |
4794 | "features"); | |
4795 | if (ret) { | |
4796 | kfree(ef); | |
4797 | goto out; | |
4798 | } | |
4799 | ||
4800 | ext4_feat = ef; | |
4801 | ret = 0; | |
4802 | out: | |
4803 | return ret; | |
4804 | } | |
4805 | ||
5dabfc78 | 4806 | static int __init ext4_init_fs(void) |
ac27a0ec | 4807 | { |
c9de560d AT |
4808 | int err; |
4809 | ||
12e9b892 | 4810 | ext4_check_flag_values(); |
5dabfc78 | 4811 | err = ext4_init_pageio(); |
6fd058f7 TT |
4812 | if (err) |
4813 | return err; | |
5dabfc78 | 4814 | err = ext4_init_system_zone(); |
bd2d0210 TT |
4815 | if (err) |
4816 | goto out5; | |
3197ebdb TT |
4817 | ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj); |
4818 | if (!ext4_kset) | |
6fd058f7 | 4819 | goto out4; |
9f6200bb | 4820 | ext4_proc_root = proc_mkdir("fs/ext4", NULL); |
857ac889 LC |
4821 | |
4822 | err = ext4_init_feat_adverts(); | |
4823 | ||
5dabfc78 | 4824 | err = ext4_init_mballoc(); |
ac27a0ec | 4825 | if (err) |
6fd058f7 | 4826 | goto out3; |
c9de560d | 4827 | |
5dabfc78 | 4828 | err = ext4_init_xattr(); |
c9de560d AT |
4829 | if (err) |
4830 | goto out2; | |
ac27a0ec DK |
4831 | err = init_inodecache(); |
4832 | if (err) | |
4833 | goto out1; | |
24b58424 TT |
4834 | register_as_ext2(); |
4835 | register_as_ext3(); | |
03010a33 | 4836 | err = register_filesystem(&ext4_fs_type); |
ac27a0ec DK |
4837 | if (err) |
4838 | goto out; | |
bfff6873 LC |
4839 | |
4840 | ext4_li_info = NULL; | |
4841 | mutex_init(&ext4_li_mtx); | |
ac27a0ec DK |
4842 | return 0; |
4843 | out: | |
24b58424 TT |
4844 | unregister_as_ext2(); |
4845 | unregister_as_ext3(); | |
ac27a0ec DK |
4846 | destroy_inodecache(); |
4847 | out1: | |
5dabfc78 | 4848 | ext4_exit_xattr(); |
c9de560d | 4849 | out2: |
5dabfc78 | 4850 | ext4_exit_mballoc(); |
6fd058f7 | 4851 | out3: |
857ac889 | 4852 | kfree(ext4_feat); |
6fd058f7 TT |
4853 | remove_proc_entry("fs/ext4", NULL); |
4854 | kset_unregister(ext4_kset); | |
4855 | out4: | |
5dabfc78 | 4856 | ext4_exit_system_zone(); |
bd2d0210 | 4857 | out5: |
5dabfc78 | 4858 | ext4_exit_pageio(); |
ac27a0ec DK |
4859 | return err; |
4860 | } | |
4861 | ||
5dabfc78 | 4862 | static void __exit ext4_exit_fs(void) |
ac27a0ec | 4863 | { |
bfff6873 | 4864 | ext4_destroy_lazyinit_thread(); |
24b58424 TT |
4865 | unregister_as_ext2(); |
4866 | unregister_as_ext3(); | |
03010a33 | 4867 | unregister_filesystem(&ext4_fs_type); |
ac27a0ec | 4868 | destroy_inodecache(); |
5dabfc78 TT |
4869 | ext4_exit_xattr(); |
4870 | ext4_exit_mballoc(); | |
9f6200bb | 4871 | remove_proc_entry("fs/ext4", NULL); |
3197ebdb | 4872 | kset_unregister(ext4_kset); |
5dabfc78 TT |
4873 | ext4_exit_system_zone(); |
4874 | ext4_exit_pageio(); | |
ac27a0ec DK |
4875 | } |
4876 | ||
4877 | MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); | |
83982b6f | 4878 | MODULE_DESCRIPTION("Fourth Extended Filesystem"); |
ac27a0ec | 4879 | MODULE_LICENSE("GPL"); |
5dabfc78 TT |
4880 | module_init(ext4_init_fs) |
4881 | module_exit(ext4_exit_fs) |