Commit | Line | Data |
---|---|---|
f5166768 | 1 | // SPDX-License-Identifier: GPL-2.0 |
ac27a0ec | 2 | /* |
617ba13b | 3 | * linux/fs/ext4/super.c |
ac27a0ec DK |
4 | * |
5 | * Copyright (C) 1992, 1993, 1994, 1995 | |
6 | * Remy Card (card@masi.ibp.fr) | |
7 | * Laboratoire MASI - Institut Blaise Pascal | |
8 | * Universite Pierre et Marie Curie (Paris VI) | |
9 | * | |
10 | * from | |
11 | * | |
12 | * linux/fs/minix/inode.c | |
13 | * | |
14 | * Copyright (C) 1991, 1992 Linus Torvalds | |
15 | * | |
16 | * Big-endian to little-endian byte-swapping/bitmaps by | |
17 | * David S. Miller (davem@caip.rutgers.edu), 1995 | |
18 | */ | |
19 | ||
20 | #include <linux/module.h> | |
21 | #include <linux/string.h> | |
22 | #include <linux/fs.h> | |
23 | #include <linux/time.h> | |
c5ca7c76 | 24 | #include <linux/vmalloc.h> |
ac27a0ec DK |
25 | #include <linux/slab.h> |
26 | #include <linux/init.h> | |
27 | #include <linux/blkdev.h> | |
66114cad | 28 | #include <linux/backing-dev.h> |
ac27a0ec | 29 | #include <linux/parser.h> |
ac27a0ec | 30 | #include <linux/buffer_head.h> |
a5694255 | 31 | #include <linux/exportfs.h> |
ac27a0ec DK |
32 | #include <linux/vfs.h> |
33 | #include <linux/random.h> | |
34 | #include <linux/mount.h> | |
35 | #include <linux/namei.h> | |
36 | #include <linux/quotaops.h> | |
37 | #include <linux/seq_file.h> | |
3197ebdb | 38 | #include <linux/ctype.h> |
1330593e | 39 | #include <linux/log2.h> |
717d50e4 | 40 | #include <linux/crc16.h> |
ef510424 | 41 | #include <linux/dax.h> |
7abc52c2 | 42 | #include <linux/cleancache.h> |
7c0f6ba6 | 43 | #include <linux/uaccess.h> |
ee73f9a5 | 44 | #include <linux/iversion.h> |
c83ad55e | 45 | #include <linux/unicode.h> |
c6a564ff | 46 | #include <linux/part_stat.h> |
bfff6873 LC |
47 | #include <linux/kthread.h> |
48 | #include <linux/freezer.h> | |
9a089b21 | 49 | #include <linux/fsnotify.h> |
e5a185c2 LC |
50 | #include <linux/fs_context.h> |
51 | #include <linux/fs_parser.h> | |
bfff6873 | 52 | |
3dcf5451 | 53 | #include "ext4.h" |
4a092d73 | 54 | #include "ext4_extents.h" /* Needed for trace points definition */ |
3dcf5451 | 55 | #include "ext4_jbd2.h" |
ac27a0ec DK |
56 | #include "xattr.h" |
57 | #include "acl.h" | |
3661d286 | 58 | #include "mballoc.h" |
0c9ec4be | 59 | #include "fsmap.h" |
ac27a0ec | 60 | |
9bffad1e TT |
61 | #define CREATE_TRACE_POINTS |
62 | #include <trace/events/ext4.h> | |
63 | ||
0b75a840 | 64 | static struct ext4_lazy_init *ext4_li_info; |
59ebc7fd | 65 | static DEFINE_MUTEX(ext4_li_mtx); |
e294a537 | 66 | static struct ratelimit_state ext4_mount_msg_ratelimit; |
9f6200bb | 67 | |
617ba13b | 68 | static int ext4_load_journal(struct super_block *, struct ext4_super_block *, |
ac27a0ec | 69 | unsigned long journal_devnum); |
2adf6da8 | 70 | static int ext4_show_options(struct seq_file *seq, struct dentry *root); |
2d01ddc8 | 71 | static void ext4_update_super(struct super_block *sb); |
4392fbc4 | 72 | static int ext4_commit_super(struct super_block *sb); |
11215630 | 73 | static int ext4_mark_recovery_complete(struct super_block *sb, |
2b2d6d01 | 74 | struct ext4_super_block *es); |
11215630 JK |
75 | static int ext4_clear_journal_err(struct super_block *sb, |
76 | struct ext4_super_block *es); | |
617ba13b | 77 | static int ext4_sync_fs(struct super_block *sb, int wait); |
2b2d6d01 TT |
78 | static int ext4_remount(struct super_block *sb, int *flags, char *data); |
79 | static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); | |
c4be0c1d | 80 | static int ext4_unfreeze(struct super_block *sb); |
c4be0c1d | 81 | static int ext4_freeze(struct super_block *sb); |
152a0836 AV |
82 | static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, |
83 | const char *dev_name, void *data); | |
2035e776 TT |
84 | static inline int ext2_feature_set_ok(struct super_block *sb); |
85 | static inline int ext3_feature_set_ok(struct super_block *sb); | |
bfff6873 LC |
86 | static void ext4_destroy_lazyinit_thread(void); |
87 | static void ext4_unregister_li_request(struct super_block *sb); | |
8f1f7453 | 88 | static void ext4_clear_request_list(void); |
c6cb7e77 EW |
89 | static struct inode *ext4_get_journal_inode(struct super_block *sb, |
90 | unsigned int journal_inum); | |
da812f61 | 91 | static int ext4_validate_options(struct fs_context *fc); |
b6bd2435 LC |
92 | static int ext4_check_opt_consistency(struct fs_context *fc, |
93 | struct super_block *sb); | |
6e47a3cc | 94 | static int ext4_apply_options(struct fs_context *fc, struct super_block *sb); |
ac27a0ec | 95 | |
e74031fd JK |
96 | /* |
97 | * Lock ordering | |
98 | * | |
e74031fd | 99 | * page fault path: |
d4f5258e JK |
100 | * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start |
101 | * -> page lock -> i_data_sem (rw) | |
e74031fd JK |
102 | * |
103 | * buffered write path: | |
c1e8d7c6 | 104 | * sb_start_write -> i_mutex -> mmap_lock |
e74031fd JK |
105 | * sb_start_write -> i_mutex -> transaction start -> page lock -> |
106 | * i_data_sem (rw) | |
107 | * | |
108 | * truncate: | |
d4f5258e JK |
109 | * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) -> |
110 | * page lock | |
111 | * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start -> | |
1d39834f | 112 | * i_data_sem (rw) |
e74031fd JK |
113 | * |
114 | * direct IO: | |
c1e8d7c6 | 115 | * sb_start_write -> i_mutex -> mmap_lock |
1d39834f | 116 | * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw) |
e74031fd JK |
117 | * |
118 | * writepages: | |
119 | * transaction start -> page lock(s) -> i_data_sem (rw) | |
120 | */ | |
121 | ||
c290ea01 | 122 | #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) |
2035e776 TT |
123 | static struct file_system_type ext2_fs_type = { |
124 | .owner = THIS_MODULE, | |
125 | .name = "ext2", | |
126 | .mount = ext4_mount, | |
127 | .kill_sb = kill_block_super, | |
128 | .fs_flags = FS_REQUIRES_DEV, | |
129 | }; | |
7f78e035 | 130 | MODULE_ALIAS_FS("ext2"); |
fa7614dd | 131 | MODULE_ALIAS("ext2"); |
2035e776 TT |
132 | #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) |
133 | #else | |
134 | #define IS_EXT2_SB(sb) (0) | |
135 | #endif | |
136 | ||
137 | ||
ba69f9ab JK |
138 | static struct file_system_type ext3_fs_type = { |
139 | .owner = THIS_MODULE, | |
140 | .name = "ext3", | |
152a0836 | 141 | .mount = ext4_mount, |
ba69f9ab JK |
142 | .kill_sb = kill_block_super, |
143 | .fs_flags = FS_REQUIRES_DEV, | |
144 | }; | |
7f78e035 | 145 | MODULE_ALIAS_FS("ext3"); |
fa7614dd | 146 | MODULE_ALIAS("ext3"); |
ba69f9ab | 147 | #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) |
bd81d8ee | 148 | |
fa491b14 | 149 | |
150 | static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags, | |
151 | bh_end_io_t *end_io) | |
152 | { | |
153 | /* | |
154 | * buffer's verified bit is no longer valid after reading from | |
155 | * disk again due to write out error, clear it to make sure we | |
156 | * recheck the buffer contents. | |
157 | */ | |
158 | clear_buffer_verified(bh); | |
159 | ||
160 | bh->b_end_io = end_io ? end_io : end_buffer_read_sync; | |
161 | get_bh(bh); | |
162 | submit_bh(REQ_OP_READ, op_flags, bh); | |
163 | } | |
164 | ||
165 | void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags, | |
166 | bh_end_io_t *end_io) | |
167 | { | |
168 | BUG_ON(!buffer_locked(bh)); | |
169 | ||
170 | if (ext4_buffer_uptodate(bh)) { | |
171 | unlock_buffer(bh); | |
172 | return; | |
173 | } | |
174 | __ext4_read_bh(bh, op_flags, end_io); | |
175 | } | |
176 | ||
177 | int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io) | |
178 | { | |
179 | BUG_ON(!buffer_locked(bh)); | |
180 | ||
181 | if (ext4_buffer_uptodate(bh)) { | |
182 | unlock_buffer(bh); | |
183 | return 0; | |
184 | } | |
185 | ||
186 | __ext4_read_bh(bh, op_flags, end_io); | |
187 | ||
188 | wait_on_buffer(bh); | |
189 | if (buffer_uptodate(bh)) | |
190 | return 0; | |
191 | return -EIO; | |
192 | } | |
193 | ||
194 | int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait) | |
195 | { | |
196 | if (trylock_buffer(bh)) { | |
197 | if (wait) | |
198 | return ext4_read_bh(bh, op_flags, NULL); | |
199 | ext4_read_bh_nowait(bh, op_flags, NULL); | |
200 | return 0; | |
201 | } | |
202 | if (wait) { | |
203 | wait_on_buffer(bh); | |
204 | if (buffer_uptodate(bh)) | |
205 | return 0; | |
206 | return -EIO; | |
207 | } | |
208 | return 0; | |
209 | } | |
210 | ||
fb265c9c | 211 | /* |
8394a6ab | 212 | * This works like __bread_gfp() except it uses ERR_PTR for error |
fb265c9c TT |
213 | * returns. Currently with sb_bread it's impossible to distinguish |
214 | * between ENOMEM and EIO situations (since both result in a NULL | |
215 | * return. | |
216 | */ | |
8394a6ab | 217 | static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb, |
218 | sector_t block, int op_flags, | |
219 | gfp_t gfp) | |
fb265c9c | 220 | { |
2d069c08 | 221 | struct buffer_head *bh; |
222 | int ret; | |
fb265c9c | 223 | |
8394a6ab | 224 | bh = sb_getblk_gfp(sb, block, gfp); |
fb265c9c TT |
225 | if (bh == NULL) |
226 | return ERR_PTR(-ENOMEM); | |
cf2834a5 | 227 | if (ext4_buffer_uptodate(bh)) |
fb265c9c | 228 | return bh; |
2d069c08 | 229 | |
230 | ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true); | |
231 | if (ret) { | |
232 | put_bh(bh); | |
233 | return ERR_PTR(ret); | |
234 | } | |
235 | return bh; | |
fb265c9c TT |
236 | } |
237 | ||
8394a6ab | 238 | struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block, |
239 | int op_flags) | |
240 | { | |
241 | return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE); | |
242 | } | |
243 | ||
244 | struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb, | |
245 | sector_t block) | |
246 | { | |
247 | return __ext4_sb_bread_gfp(sb, block, 0, 0); | |
248 | } | |
249 | ||
5df1d412 | 250 | void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block) |
251 | { | |
252 | struct buffer_head *bh = sb_getblk_gfp(sb, block, 0); | |
253 | ||
254 | if (likely(bh)) { | |
255 | ext4_read_bh_lock(bh, REQ_RAHEAD, false); | |
256 | brelse(bh); | |
257 | } | |
fb265c9c TT |
258 | } |
259 | ||
d25425f8 DW |
260 | static int ext4_verify_csum_type(struct super_block *sb, |
261 | struct ext4_super_block *es) | |
262 | { | |
e2b911c5 | 263 | if (!ext4_has_feature_metadata_csum(sb)) |
d25425f8 DW |
264 | return 1; |
265 | ||
266 | return es->s_checksum_type == EXT4_CRC32C_CHKSUM; | |
267 | } | |
268 | ||
a9c47317 DW |
269 | static __le32 ext4_superblock_csum(struct super_block *sb, |
270 | struct ext4_super_block *es) | |
271 | { | |
272 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
273 | int offset = offsetof(struct ext4_super_block, s_checksum); | |
274 | __u32 csum; | |
275 | ||
276 | csum = ext4_chksum(sbi, ~0, (char *)es, offset); | |
277 | ||
278 | return cpu_to_le32(csum); | |
279 | } | |
280 | ||
c197855e SH |
281 | static int ext4_superblock_csum_verify(struct super_block *sb, |
282 | struct ext4_super_block *es) | |
a9c47317 | 283 | { |
9aa5d32b | 284 | if (!ext4_has_metadata_csum(sb)) |
a9c47317 DW |
285 | return 1; |
286 | ||
287 | return es->s_checksum == ext4_superblock_csum(sb, es); | |
288 | } | |
289 | ||
06db49e6 | 290 | void ext4_superblock_csum_set(struct super_block *sb) |
a9c47317 | 291 | { |
06db49e6 TT |
292 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
293 | ||
9aa5d32b | 294 | if (!ext4_has_metadata_csum(sb)) |
a9c47317 DW |
295 | return; |
296 | ||
297 | es->s_checksum = ext4_superblock_csum(sb, es); | |
298 | } | |
299 | ||
8fadc143 AR |
300 | ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, |
301 | struct ext4_group_desc *bg) | |
bd81d8ee | 302 | { |
3a14589c | 303 | return le32_to_cpu(bg->bg_block_bitmap_lo) | |
8fadc143 | 304 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 305 | (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); |
bd81d8ee LV |
306 | } |
307 | ||
8fadc143 AR |
308 | ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, |
309 | struct ext4_group_desc *bg) | |
bd81d8ee | 310 | { |
5272f837 | 311 | return le32_to_cpu(bg->bg_inode_bitmap_lo) | |
8fadc143 | 312 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 313 | (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); |
bd81d8ee LV |
314 | } |
315 | ||
8fadc143 AR |
316 | ext4_fsblk_t ext4_inode_table(struct super_block *sb, |
317 | struct ext4_group_desc *bg) | |
bd81d8ee | 318 | { |
5272f837 | 319 | return le32_to_cpu(bg->bg_inode_table_lo) | |
8fadc143 | 320 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? |
0b8e58a1 | 321 | (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); |
bd81d8ee LV |
322 | } |
323 | ||
021b65bb TT |
324 | __u32 ext4_free_group_clusters(struct super_block *sb, |
325 | struct ext4_group_desc *bg) | |
560671a0 AK |
326 | { |
327 | return le16_to_cpu(bg->bg_free_blocks_count_lo) | | |
328 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 329 | (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); |
560671a0 AK |
330 | } |
331 | ||
332 | __u32 ext4_free_inodes_count(struct super_block *sb, | |
333 | struct ext4_group_desc *bg) | |
334 | { | |
335 | return le16_to_cpu(bg->bg_free_inodes_count_lo) | | |
336 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 337 | (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); |
560671a0 AK |
338 | } |
339 | ||
340 | __u32 ext4_used_dirs_count(struct super_block *sb, | |
341 | struct ext4_group_desc *bg) | |
342 | { | |
343 | return le16_to_cpu(bg->bg_used_dirs_count_lo) | | |
344 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 345 | (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); |
560671a0 AK |
346 | } |
347 | ||
348 | __u32 ext4_itable_unused_count(struct super_block *sb, | |
349 | struct ext4_group_desc *bg) | |
350 | { | |
351 | return le16_to_cpu(bg->bg_itable_unused_lo) | | |
352 | (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
0b8e58a1 | 353 | (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); |
560671a0 AK |
354 | } |
355 | ||
8fadc143 AR |
356 | void ext4_block_bitmap_set(struct super_block *sb, |
357 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 358 | { |
3a14589c | 359 | bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
360 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
361 | bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
362 | } |
363 | ||
8fadc143 AR |
364 | void ext4_inode_bitmap_set(struct super_block *sb, |
365 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 366 | { |
5272f837 | 367 | bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
368 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
369 | bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
370 | } |
371 | ||
8fadc143 AR |
372 | void ext4_inode_table_set(struct super_block *sb, |
373 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
bd81d8ee | 374 | { |
5272f837 | 375 | bg->bg_inode_table_lo = cpu_to_le32((u32)blk); |
8fadc143 AR |
376 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) |
377 | bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); | |
bd81d8ee LV |
378 | } |
379 | ||
021b65bb TT |
380 | void ext4_free_group_clusters_set(struct super_block *sb, |
381 | struct ext4_group_desc *bg, __u32 count) | |
560671a0 AK |
382 | { |
383 | bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); | |
384 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
385 | bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); | |
386 | } | |
387 | ||
388 | void ext4_free_inodes_set(struct super_block *sb, | |
389 | struct ext4_group_desc *bg, __u32 count) | |
390 | { | |
391 | bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); | |
392 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
393 | bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); | |
394 | } | |
395 | ||
396 | void ext4_used_dirs_set(struct super_block *sb, | |
397 | struct ext4_group_desc *bg, __u32 count) | |
398 | { | |
399 | bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); | |
400 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
401 | bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); | |
402 | } | |
403 | ||
404 | void ext4_itable_unused_set(struct super_block *sb, | |
405 | struct ext4_group_desc *bg, __u32 count) | |
406 | { | |
407 | bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); | |
408 | if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
409 | bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); | |
410 | } | |
411 | ||
c92dc856 | 412 | static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now) |
6a0678a7 | 413 | { |
6a0678a7 AB |
414 | now = clamp_val(now, 0, (1ull << 40) - 1); |
415 | ||
416 | *lo = cpu_to_le32(lower_32_bits(now)); | |
417 | *hi = upper_32_bits(now); | |
418 | } | |
419 | ||
420 | static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi) | |
421 | { | |
422 | return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo); | |
423 | } | |
424 | #define ext4_update_tstamp(es, tstamp) \ | |
c92dc856 JK |
425 | __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \ |
426 | ktime_get_real_seconds()) | |
6a0678a7 AB |
427 | #define ext4_get_tstamp(es, tstamp) \ |
428 | __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) | |
d3d1faf6 | 429 | |
bdfe0cbd TT |
430 | /* |
431 | * The del_gendisk() function uninitializes the disk-specific data | |
432 | * structures, including the bdi structure, without telling anyone | |
433 | * else. Once this happens, any attempt to call mark_buffer_dirty() | |
434 | * (for example, by ext4_commit_super), will cause a kernel OOPS. | |
435 | * This is a kludge to prevent these oops until we can put in a proper | |
436 | * hook in del_gendisk() to inform the VFS and file system layers. | |
437 | */ | |
438 | static int block_device_ejected(struct super_block *sb) | |
439 | { | |
440 | struct inode *bd_inode = sb->s_bdev->bd_inode; | |
441 | struct backing_dev_info *bdi = inode_to_bdi(bd_inode); | |
442 | ||
443 | return bdi->dev == NULL; | |
444 | } | |
445 | ||
18aadd47 BJ |
446 | static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) |
447 | { | |
448 | struct super_block *sb = journal->j_private; | |
449 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
450 | int error = is_journal_aborted(journal); | |
5d3ee208 | 451 | struct ext4_journal_cb_entry *jce; |
18aadd47 | 452 | |
5d3ee208 | 453 | BUG_ON(txn->t_state == T_FINISHED); |
a0154344 DJ |
454 | |
455 | ext4_process_freed_data(sb, txn->t_tid); | |
456 | ||
18aadd47 | 457 | spin_lock(&sbi->s_md_lock); |
5d3ee208 DM |
458 | while (!list_empty(&txn->t_private_list)) { |
459 | jce = list_entry(txn->t_private_list.next, | |
460 | struct ext4_journal_cb_entry, jce_list); | |
18aadd47 BJ |
461 | list_del_init(&jce->jce_list); |
462 | spin_unlock(&sbi->s_md_lock); | |
463 | jce->jce_func(sb, jce, error); | |
464 | spin_lock(&sbi->s_md_lock); | |
465 | } | |
466 | spin_unlock(&sbi->s_md_lock); | |
467 | } | |
1c13d5c0 | 468 | |
afb585a9 MFO |
469 | /* |
470 | * This writepage callback for write_cache_pages() | |
471 | * takes care of a few cases after page cleaning. | |
472 | * | |
473 | * write_cache_pages() already checks for dirty pages | |
474 | * and calls clear_page_dirty_for_io(), which we want, | |
475 | * to write protect the pages. | |
476 | * | |
477 | * However, we may have to redirty a page (see below.) | |
478 | */ | |
479 | static int ext4_journalled_writepage_callback(struct page *page, | |
480 | struct writeback_control *wbc, | |
481 | void *data) | |
482 | { | |
483 | transaction_t *transaction = (transaction_t *) data; | |
484 | struct buffer_head *bh, *head; | |
485 | struct journal_head *jh; | |
486 | ||
487 | bh = head = page_buffers(page); | |
488 | do { | |
489 | /* | |
490 | * We have to redirty a page in these cases: | |
491 | * 1) If buffer is dirty, it means the page was dirty because it | |
492 | * contains a buffer that needs checkpointing. So the dirty bit | |
493 | * needs to be preserved so that checkpointing writes the buffer | |
494 | * properly. | |
495 | * 2) If buffer is not part of the committing transaction | |
496 | * (we may have just accidentally come across this buffer because | |
497 | * inode range tracking is not exact) or if the currently running | |
498 | * transaction already contains this buffer as well, dirty bit | |
499 | * needs to be preserved so that the buffer gets writeprotected | |
500 | * properly on running transaction's commit. | |
501 | */ | |
502 | jh = bh2jh(bh); | |
503 | if (buffer_dirty(bh) || | |
504 | (jh && (jh->b_transaction != transaction || | |
505 | jh->b_next_transaction))) { | |
506 | redirty_page_for_writepage(wbc, page); | |
507 | goto out; | |
508 | } | |
509 | } while ((bh = bh->b_this_page) != head); | |
510 | ||
511 | out: | |
512 | return AOP_WRITEPAGE_ACTIVATE; | |
513 | } | |
514 | ||
515 | static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode) | |
516 | { | |
517 | struct address_space *mapping = jinode->i_vfs_inode->i_mapping; | |
518 | struct writeback_control wbc = { | |
519 | .sync_mode = WB_SYNC_ALL, | |
520 | .nr_to_write = LONG_MAX, | |
521 | .range_start = jinode->i_dirty_start, | |
522 | .range_end = jinode->i_dirty_end, | |
523 | }; | |
524 | ||
525 | return write_cache_pages(mapping, &wbc, | |
526 | ext4_journalled_writepage_callback, | |
527 | jinode->i_transaction); | |
528 | } | |
529 | ||
530 | static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode) | |
531 | { | |
532 | int ret; | |
533 | ||
534 | if (ext4_should_journal_data(jinode->i_vfs_inode)) | |
535 | ret = ext4_journalled_submit_inode_data_buffers(jinode); | |
536 | else | |
537 | ret = jbd2_journal_submit_inode_data_buffers(jinode); | |
538 | ||
539 | return ret; | |
540 | } | |
541 | ||
542 | static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode) | |
543 | { | |
544 | int ret = 0; | |
545 | ||
546 | if (!ext4_should_journal_data(jinode->i_vfs_inode)) | |
547 | ret = jbd2_journal_finish_inode_data_buffers(jinode); | |
548 | ||
549 | return ret; | |
550 | } | |
551 | ||
1dc1097f JK |
552 | static bool system_going_down(void) |
553 | { | |
554 | return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF | |
555 | || system_state == SYSTEM_RESTART; | |
556 | } | |
557 | ||
02a7780e JK |
558 | struct ext4_err_translation { |
559 | int code; | |
560 | int errno; | |
561 | }; | |
562 | ||
563 | #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err } | |
564 | ||
565 | static struct ext4_err_translation err_translation[] = { | |
566 | EXT4_ERR_TRANSLATE(EIO), | |
567 | EXT4_ERR_TRANSLATE(ENOMEM), | |
568 | EXT4_ERR_TRANSLATE(EFSBADCRC), | |
569 | EXT4_ERR_TRANSLATE(EFSCORRUPTED), | |
570 | EXT4_ERR_TRANSLATE(ENOSPC), | |
571 | EXT4_ERR_TRANSLATE(ENOKEY), | |
572 | EXT4_ERR_TRANSLATE(EROFS), | |
573 | EXT4_ERR_TRANSLATE(EFBIG), | |
574 | EXT4_ERR_TRANSLATE(EEXIST), | |
575 | EXT4_ERR_TRANSLATE(ERANGE), | |
576 | EXT4_ERR_TRANSLATE(EOVERFLOW), | |
577 | EXT4_ERR_TRANSLATE(EBUSY), | |
578 | EXT4_ERR_TRANSLATE(ENOTDIR), | |
579 | EXT4_ERR_TRANSLATE(ENOTEMPTY), | |
580 | EXT4_ERR_TRANSLATE(ESHUTDOWN), | |
581 | EXT4_ERR_TRANSLATE(EFAULT), | |
582 | }; | |
583 | ||
584 | static int ext4_errno_to_code(int errno) | |
585 | { | |
586 | int i; | |
587 | ||
588 | for (i = 0; i < ARRAY_SIZE(err_translation); i++) | |
589 | if (err_translation[i].errno == errno) | |
590 | return err_translation[i].code; | |
591 | return EXT4_ERR_UNKNOWN; | |
592 | } | |
593 | ||
2d01ddc8 JK |
594 | static void save_error_info(struct super_block *sb, int error, |
595 | __u32 ino, __u64 block, | |
596 | const char *func, unsigned int line) | |
40676623 | 597 | { |
c92dc856 | 598 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
40676623 | 599 | |
02a7780e JK |
600 | /* We default to EFSCORRUPTED error... */ |
601 | if (error == 0) | |
602 | error = EFSCORRUPTED; | |
c92dc856 JK |
603 | |
604 | spin_lock(&sbi->s_error_lock); | |
605 | sbi->s_add_error_count++; | |
606 | sbi->s_last_error_code = error; | |
607 | sbi->s_last_error_line = line; | |
608 | sbi->s_last_error_ino = ino; | |
609 | sbi->s_last_error_block = block; | |
610 | sbi->s_last_error_func = func; | |
611 | sbi->s_last_error_time = ktime_get_real_seconds(); | |
612 | if (!sbi->s_first_error_time) { | |
613 | sbi->s_first_error_code = error; | |
614 | sbi->s_first_error_line = line; | |
615 | sbi->s_first_error_ino = ino; | |
616 | sbi->s_first_error_block = block; | |
617 | sbi->s_first_error_func = func; | |
618 | sbi->s_first_error_time = sbi->s_last_error_time; | |
619 | } | |
620 | spin_unlock(&sbi->s_error_lock); | |
40676623 JK |
621 | } |
622 | ||
ac27a0ec DK |
623 | /* Deal with the reporting of failure conditions on a filesystem such as |
624 | * inconsistencies detected or read IO failures. | |
625 | * | |
626 | * On ext2, we can store the error state of the filesystem in the | |
617ba13b | 627 | * superblock. That is not possible on ext4, because we may have other |
ac27a0ec DK |
628 | * write ordering constraints on the superblock which prevent us from |
629 | * writing it out straight away; and given that the journal is about to | |
630 | * be aborted, we can't rely on the current, or future, transactions to | |
631 | * write out the superblock safely. | |
632 | * | |
dab291af | 633 | * We'll just use the jbd2_journal_abort() error code to record an error in |
d6b198bc | 634 | * the journal instead. On recovery, the journal will complain about |
ac27a0ec | 635 | * that error until we've noted it down and cleared it. |
014c9caa JK |
636 | * |
637 | * If force_ro is set, we unconditionally force the filesystem into an | |
638 | * ABORT|READONLY state, unless the error response on the fs has been set to | |
639 | * panic in which case we take the easy way out and panic immediately. This is | |
640 | * used to deal with unrecoverable failures such as journal IO errors or ENOMEM | |
641 | * at a critical moment in log management. | |
ac27a0ec | 642 | */ |
e789ca0c JK |
643 | static void ext4_handle_error(struct super_block *sb, bool force_ro, int error, |
644 | __u32 ino, __u64 block, | |
645 | const char *func, unsigned int line) | |
ac27a0ec | 646 | { |
b08070ec | 647 | journal_t *journal = EXT4_SB(sb)->s_journal; |
2d01ddc8 | 648 | bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT); |
b08070ec | 649 | |
e789ca0c | 650 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; |
327eaf73 TT |
651 | if (test_opt(sb, WARN_ON_ERROR)) |
652 | WARN_ON_ONCE(1); | |
653 | ||
2d01ddc8 JK |
654 | if (!continue_fs && !sb_rdonly(sb)) { |
655 | ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED); | |
656 | if (journal) | |
657 | jbd2_journal_abort(journal, -EIO); | |
658 | } | |
659 | ||
660 | if (!bdev_read_only(sb->s_bdev)) { | |
e789ca0c | 661 | save_error_info(sb, error, ino, block, func, line); |
2d01ddc8 JK |
662 | /* |
663 | * In case the fs should keep running, we need to writeout | |
664 | * superblock through the journal. Due to lock ordering | |
665 | * constraints, it may not be safe to do it right here so we | |
666 | * defer superblock flushing to a workqueue. | |
667 | */ | |
bb9464e0 | 668 | if (continue_fs && journal) |
2d01ddc8 JK |
669 | schedule_work(&EXT4_SB(sb)->s_error_work); |
670 | else | |
671 | ext4_commit_super(sb); | |
672 | } | |
e789ca0c | 673 | |
1dc1097f JK |
674 | /* |
675 | * We force ERRORS_RO behavior when system is rebooting. Otherwise we | |
676 | * could panic during 'reboot -f' as the underlying device got already | |
677 | * disabled. | |
678 | */ | |
014c9caa | 679 | if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) { |
617ba13b | 680 | panic("EXT4-fs (device %s): panic forced after error\n", |
ac27a0ec | 681 | sb->s_id); |
4327ba52 | 682 | } |
ac2f7ca5 YB |
683 | |
684 | if (sb_rdonly(sb) || continue_fs) | |
685 | return; | |
686 | ||
014c9caa JK |
687 | ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); |
688 | /* | |
689 | * Make sure updated value of ->s_mount_flags will be visible before | |
690 | * ->s_flags update | |
691 | */ | |
692 | smp_wmb(); | |
693 | sb->s_flags |= SB_RDONLY; | |
ac27a0ec DK |
694 | } |
695 | ||
c92dc856 JK |
696 | static void flush_stashed_error_work(struct work_struct *work) |
697 | { | |
698 | struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info, | |
699 | s_error_work); | |
2d01ddc8 JK |
700 | journal_t *journal = sbi->s_journal; |
701 | handle_t *handle; | |
c92dc856 | 702 | |
2d01ddc8 JK |
703 | /* |
704 | * If the journal is still running, we have to write out superblock | |
705 | * through the journal to avoid collisions of other journalled sb | |
706 | * updates. | |
707 | * | |
708 | * We use directly jbd2 functions here to avoid recursing back into | |
709 | * ext4 error handling code during handling of previous errors. | |
710 | */ | |
711 | if (!sb_rdonly(sbi->s_sb) && journal) { | |
558d6450 | 712 | struct buffer_head *sbh = sbi->s_sbh; |
2d01ddc8 JK |
713 | handle = jbd2_journal_start(journal, 1); |
714 | if (IS_ERR(handle)) | |
715 | goto write_directly; | |
558d6450 | 716 | if (jbd2_journal_get_write_access(handle, sbh)) { |
2d01ddc8 JK |
717 | jbd2_journal_stop(handle); |
718 | goto write_directly; | |
719 | } | |
720 | ext4_update_super(sbi->s_sb); | |
558d6450 YB |
721 | if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { |
722 | ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to " | |
723 | "superblock detected"); | |
724 | clear_buffer_write_io_error(sbh); | |
725 | set_buffer_uptodate(sbh); | |
726 | } | |
727 | ||
728 | if (jbd2_journal_dirty_metadata(handle, sbh)) { | |
2d01ddc8 JK |
729 | jbd2_journal_stop(handle); |
730 | goto write_directly; | |
731 | } | |
732 | jbd2_journal_stop(handle); | |
d578b994 | 733 | ext4_notify_error_sysfs(sbi); |
2d01ddc8 JK |
734 | return; |
735 | } | |
736 | write_directly: | |
737 | /* | |
738 | * Write through journal failed. Write sb directly to get error info | |
739 | * out and hope for the best. | |
740 | */ | |
4392fbc4 | 741 | ext4_commit_super(sbi->s_sb); |
d578b994 | 742 | ext4_notify_error_sysfs(sbi); |
ac27a0ec DK |
743 | } |
744 | ||
efbed4dc TT |
745 | #define ext4_error_ratelimit(sb) \ |
746 | ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \ | |
747 | "EXT4-fs error") | |
748 | ||
12062ddd | 749 | void __ext4_error(struct super_block *sb, const char *function, |
014c9caa | 750 | unsigned int line, bool force_ro, int error, __u64 block, |
54d3adbc | 751 | const char *fmt, ...) |
ac27a0ec | 752 | { |
0ff2ea7d | 753 | struct va_format vaf; |
ac27a0ec DK |
754 | va_list args; |
755 | ||
0db1ff22 TT |
756 | if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) |
757 | return; | |
758 | ||
ccf0f32a | 759 | trace_ext4_error(sb, function, line); |
efbed4dc TT |
760 | if (ext4_error_ratelimit(sb)) { |
761 | va_start(args, fmt); | |
762 | vaf.fmt = fmt; | |
763 | vaf.va = &args; | |
764 | printk(KERN_CRIT | |
765 | "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", | |
766 | sb->s_id, function, line, current->comm, &vaf); | |
767 | va_end(args); | |
768 | } | |
9a089b21 GKB |
769 | fsnotify_sb_error(sb, NULL, error ? error : EFSCORRUPTED); |
770 | ||
e789ca0c | 771 | ext4_handle_error(sb, force_ro, error, 0, block, function, line); |
ac27a0ec DK |
772 | } |
773 | ||
e7c96e8e | 774 | void __ext4_error_inode(struct inode *inode, const char *function, |
54d3adbc | 775 | unsigned int line, ext4_fsblk_t block, int error, |
e7c96e8e | 776 | const char *fmt, ...) |
273df556 FM |
777 | { |
778 | va_list args; | |
f7c21177 | 779 | struct va_format vaf; |
273df556 | 780 | |
0db1ff22 TT |
781 | if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) |
782 | return; | |
783 | ||
ccf0f32a | 784 | trace_ext4_error(inode->i_sb, function, line); |
efbed4dc TT |
785 | if (ext4_error_ratelimit(inode->i_sb)) { |
786 | va_start(args, fmt); | |
787 | vaf.fmt = fmt; | |
788 | vaf.va = &args; | |
789 | if (block) | |
790 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " | |
791 | "inode #%lu: block %llu: comm %s: %pV\n", | |
792 | inode->i_sb->s_id, function, line, inode->i_ino, | |
793 | block, current->comm, &vaf); | |
794 | else | |
795 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " | |
796 | "inode #%lu: comm %s: %pV\n", | |
797 | inode->i_sb->s_id, function, line, inode->i_ino, | |
798 | current->comm, &vaf); | |
799 | va_end(args); | |
800 | } | |
9a089b21 GKB |
801 | fsnotify_sb_error(inode->i_sb, inode, error ? error : EFSCORRUPTED); |
802 | ||
e789ca0c JK |
803 | ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block, |
804 | function, line); | |
273df556 FM |
805 | } |
806 | ||
e7c96e8e JP |
807 | void __ext4_error_file(struct file *file, const char *function, |
808 | unsigned int line, ext4_fsblk_t block, | |
809 | const char *fmt, ...) | |
273df556 FM |
810 | { |
811 | va_list args; | |
f7c21177 | 812 | struct va_format vaf; |
496ad9aa | 813 | struct inode *inode = file_inode(file); |
273df556 FM |
814 | char pathname[80], *path; |
815 | ||
0db1ff22 TT |
816 | if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) |
817 | return; | |
818 | ||
ccf0f32a | 819 | trace_ext4_error(inode->i_sb, function, line); |
efbed4dc | 820 | if (ext4_error_ratelimit(inode->i_sb)) { |
9bf39ab2 | 821 | path = file_path(file, pathname, sizeof(pathname)); |
efbed4dc TT |
822 | if (IS_ERR(path)) |
823 | path = "(unknown)"; | |
824 | va_start(args, fmt); | |
825 | vaf.fmt = fmt; | |
826 | vaf.va = &args; | |
827 | if (block) | |
828 | printk(KERN_CRIT | |
829 | "EXT4-fs error (device %s): %s:%d: inode #%lu: " | |
830 | "block %llu: comm %s: path %s: %pV\n", | |
831 | inode->i_sb->s_id, function, line, inode->i_ino, | |
832 | block, current->comm, path, &vaf); | |
833 | else | |
834 | printk(KERN_CRIT | |
835 | "EXT4-fs error (device %s): %s:%d: inode #%lu: " | |
836 | "comm %s: path %s: %pV\n", | |
837 | inode->i_sb->s_id, function, line, inode->i_ino, | |
838 | current->comm, path, &vaf); | |
839 | va_end(args); | |
840 | } | |
9a089b21 GKB |
841 | fsnotify_sb_error(inode->i_sb, inode, EFSCORRUPTED); |
842 | ||
e789ca0c JK |
843 | ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block, |
844 | function, line); | |
273df556 FM |
845 | } |
846 | ||
722887dd TT |
847 | const char *ext4_decode_error(struct super_block *sb, int errno, |
848 | char nbuf[16]) | |
ac27a0ec DK |
849 | { |
850 | char *errstr = NULL; | |
851 | ||
852 | switch (errno) { | |
6a797d27 DW |
853 | case -EFSCORRUPTED: |
854 | errstr = "Corrupt filesystem"; | |
855 | break; | |
856 | case -EFSBADCRC: | |
857 | errstr = "Filesystem failed CRC"; | |
858 | break; | |
ac27a0ec DK |
859 | case -EIO: |
860 | errstr = "IO failure"; | |
861 | break; | |
862 | case -ENOMEM: | |
863 | errstr = "Out of memory"; | |
864 | break; | |
865 | case -EROFS: | |
78f1ddbb TT |
866 | if (!sb || (EXT4_SB(sb)->s_journal && |
867 | EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) | |
ac27a0ec DK |
868 | errstr = "Journal has aborted"; |
869 | else | |
870 | errstr = "Readonly filesystem"; | |
871 | break; | |
872 | default: | |
873 | /* If the caller passed in an extra buffer for unknown | |
874 | * errors, textualise them now. Else we just return | |
875 | * NULL. */ | |
876 | if (nbuf) { | |
877 | /* Check for truncated error codes... */ | |
878 | if (snprintf(nbuf, 16, "error %d", -errno) >= 0) | |
879 | errstr = nbuf; | |
880 | } | |
881 | break; | |
882 | } | |
883 | ||
884 | return errstr; | |
885 | } | |
886 | ||
617ba13b | 887 | /* __ext4_std_error decodes expected errors from journaling functions |
ac27a0ec DK |
888 | * automatically and invokes the appropriate error response. */ |
889 | ||
c398eda0 TT |
890 | void __ext4_std_error(struct super_block *sb, const char *function, |
891 | unsigned int line, int errno) | |
ac27a0ec DK |
892 | { |
893 | char nbuf[16]; | |
894 | const char *errstr; | |
895 | ||
0db1ff22 TT |
896 | if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) |
897 | return; | |
898 | ||
ac27a0ec DK |
899 | /* Special case: if the error is EROFS, and we're not already |
900 | * inside a transaction, then there's really no point in logging | |
901 | * an error. */ | |
bc98a42c | 902 | if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb)) |
ac27a0ec DK |
903 | return; |
904 | ||
efbed4dc TT |
905 | if (ext4_error_ratelimit(sb)) { |
906 | errstr = ext4_decode_error(sb, errno, nbuf); | |
907 | printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", | |
908 | sb->s_id, function, line, errstr); | |
909 | } | |
9a089b21 | 910 | fsnotify_sb_error(sb, NULL, errno ? errno : EFSCORRUPTED); |
ac27a0ec | 911 | |
e789ca0c | 912 | ext4_handle_error(sb, false, -errno, 0, 0, function, line); |
ac27a0ec DK |
913 | } |
914 | ||
e7c96e8e JP |
915 | void __ext4_msg(struct super_block *sb, |
916 | const char *prefix, const char *fmt, ...) | |
b31e1552 | 917 | { |
0ff2ea7d | 918 | struct va_format vaf; |
b31e1552 ES |
919 | va_list args; |
920 | ||
da812f61 LC |
921 | if (sb) { |
922 | atomic_inc(&EXT4_SB(sb)->s_msg_count); | |
923 | if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), | |
924 | "EXT4-fs")) | |
925 | return; | |
926 | } | |
efbed4dc | 927 | |
b31e1552 | 928 | va_start(args, fmt); |
0ff2ea7d JP |
929 | vaf.fmt = fmt; |
930 | vaf.va = &args; | |
da812f61 LC |
931 | if (sb) |
932 | printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); | |
933 | else | |
934 | printk("%sEXT4-fs: %pV\n", prefix, &vaf); | |
b31e1552 ES |
935 | va_end(args); |
936 | } | |
937 | ||
1cf006ed DM |
938 | static int ext4_warning_ratelimit(struct super_block *sb) |
939 | { | |
940 | atomic_inc(&EXT4_SB(sb)->s_warning_count); | |
941 | return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), | |
942 | "EXT4-fs warning"); | |
943 | } | |
b03a2f7e | 944 | |
12062ddd | 945 | void __ext4_warning(struct super_block *sb, const char *function, |
c398eda0 | 946 | unsigned int line, const char *fmt, ...) |
ac27a0ec | 947 | { |
0ff2ea7d | 948 | struct va_format vaf; |
ac27a0ec DK |
949 | va_list args; |
950 | ||
b03a2f7e | 951 | if (!ext4_warning_ratelimit(sb)) |
efbed4dc TT |
952 | return; |
953 | ||
ac27a0ec | 954 | va_start(args, fmt); |
0ff2ea7d JP |
955 | vaf.fmt = fmt; |
956 | vaf.va = &args; | |
957 | printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", | |
958 | sb->s_id, function, line, &vaf); | |
ac27a0ec DK |
959 | va_end(args); |
960 | } | |
961 | ||
b03a2f7e AD |
962 | void __ext4_warning_inode(const struct inode *inode, const char *function, |
963 | unsigned int line, const char *fmt, ...) | |
964 | { | |
965 | struct va_format vaf; | |
966 | va_list args; | |
967 | ||
968 | if (!ext4_warning_ratelimit(inode->i_sb)) | |
969 | return; | |
970 | ||
971 | va_start(args, fmt); | |
972 | vaf.fmt = fmt; | |
973 | vaf.va = &args; | |
974 | printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: " | |
975 | "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id, | |
976 | function, line, inode->i_ino, current->comm, &vaf); | |
977 | va_end(args); | |
978 | } | |
979 | ||
e29136f8 TT |
980 | void __ext4_grp_locked_error(const char *function, unsigned int line, |
981 | struct super_block *sb, ext4_group_t grp, | |
982 | unsigned long ino, ext4_fsblk_t block, | |
983 | const char *fmt, ...) | |
5d1b1b3f AK |
984 | __releases(bitlock) |
985 | __acquires(bitlock) | |
986 | { | |
0ff2ea7d | 987 | struct va_format vaf; |
5d1b1b3f | 988 | va_list args; |
5d1b1b3f | 989 | |
0db1ff22 TT |
990 | if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) |
991 | return; | |
992 | ||
ccf0f32a | 993 | trace_ext4_error(sb, function, line); |
efbed4dc TT |
994 | if (ext4_error_ratelimit(sb)) { |
995 | va_start(args, fmt); | |
996 | vaf.fmt = fmt; | |
997 | vaf.va = &args; | |
998 | printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", | |
999 | sb->s_id, function, line, grp); | |
1000 | if (ino) | |
1001 | printk(KERN_CONT "inode %lu: ", ino); | |
1002 | if (block) | |
1003 | printk(KERN_CONT "block %llu:", | |
1004 | (unsigned long long) block); | |
1005 | printk(KERN_CONT "%pV\n", &vaf); | |
1006 | va_end(args); | |
1007 | } | |
5d1b1b3f AK |
1008 | |
1009 | if (test_opt(sb, ERRORS_CONT)) { | |
c92dc856 JK |
1010 | if (test_opt(sb, WARN_ON_ERROR)) |
1011 | WARN_ON_ONCE(1); | |
e789ca0c | 1012 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; |
2d01ddc8 JK |
1013 | if (!bdev_read_only(sb->s_bdev)) { |
1014 | save_error_info(sb, EFSCORRUPTED, ino, block, function, | |
1015 | line); | |
e789ca0c | 1016 | schedule_work(&EXT4_SB(sb)->s_error_work); |
2d01ddc8 | 1017 | } |
5d1b1b3f AK |
1018 | return; |
1019 | } | |
1020 | ext4_unlock_group(sb, grp); | |
e789ca0c | 1021 | ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line); |
5d1b1b3f AK |
1022 | /* |
1023 | * We only get here in the ERRORS_RO case; relocking the group | |
1024 | * may be dangerous, but nothing bad will happen since the | |
1025 | * filesystem will have already been marked read/only and the | |
1026 | * journal has been aborted. We return 1 as a hint to callers | |
1027 | * who might what to use the return value from | |
25985edc | 1028 | * ext4_grp_locked_error() to distinguish between the |
5d1b1b3f AK |
1029 | * ERRORS_CONT and ERRORS_RO case, and perhaps return more |
1030 | * aggressively from the ext4 function in question, with a | |
1031 | * more appropriate error code. | |
1032 | */ | |
1033 | ext4_lock_group(sb, grp); | |
1034 | return; | |
1035 | } | |
1036 | ||
db79e6d1 WS |
1037 | void ext4_mark_group_bitmap_corrupted(struct super_block *sb, |
1038 | ext4_group_t group, | |
1039 | unsigned int flags) | |
1040 | { | |
1041 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
1042 | struct ext4_group_info *grp = ext4_get_group_info(sb, group); | |
1043 | struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL); | |
9af0b3d1 WS |
1044 | int ret; |
1045 | ||
1046 | if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) { | |
1047 | ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, | |
1048 | &grp->bb_state); | |
1049 | if (!ret) | |
1050 | percpu_counter_sub(&sbi->s_freeclusters_counter, | |
1051 | grp->bb_free); | |
db79e6d1 WS |
1052 | } |
1053 | ||
9af0b3d1 WS |
1054 | if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) { |
1055 | ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, | |
1056 | &grp->bb_state); | |
1057 | if (!ret && gdp) { | |
db79e6d1 WS |
1058 | int count; |
1059 | ||
1060 | count = ext4_free_inodes_count(sb, gdp); | |
1061 | percpu_counter_sub(&sbi->s_freeinodes_counter, | |
1062 | count); | |
1063 | } | |
db79e6d1 WS |
1064 | } |
1065 | } | |
1066 | ||
617ba13b | 1067 | void ext4_update_dynamic_rev(struct super_block *sb) |
ac27a0ec | 1068 | { |
617ba13b | 1069 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; |
ac27a0ec | 1070 | |
617ba13b | 1071 | if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) |
ac27a0ec DK |
1072 | return; |
1073 | ||
12062ddd | 1074 | ext4_warning(sb, |
ac27a0ec DK |
1075 | "updating to rev %d because of new feature flag, " |
1076 | "running e2fsck is recommended", | |
617ba13b | 1077 | EXT4_DYNAMIC_REV); |
ac27a0ec | 1078 | |
617ba13b MC |
1079 | es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); |
1080 | es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); | |
1081 | es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); | |
ac27a0ec DK |
1082 | /* leave es->s_feature_*compat flags alone */ |
1083 | /* es->s_uuid will be set by e2fsck if empty */ | |
1084 | ||
1085 | /* | |
1086 | * The rest of the superblock fields should be zero, and if not it | |
1087 | * means they are likely already in use, so leave them alone. We | |
1088 | * can leave it up to e2fsck to clean up any inconsistencies there. | |
1089 | */ | |
1090 | } | |
1091 | ||
1092 | /* | |
1093 | * Open the external journal device | |
1094 | */ | |
b31e1552 | 1095 | static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) |
ac27a0ec DK |
1096 | { |
1097 | struct block_device *bdev; | |
ac27a0ec | 1098 | |
d4d77629 | 1099 | bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); |
ac27a0ec DK |
1100 | if (IS_ERR(bdev)) |
1101 | goto fail; | |
1102 | return bdev; | |
1103 | ||
1104 | fail: | |
ea3edd4d CH |
1105 | ext4_msg(sb, KERN_ERR, |
1106 | "failed to open journal device unknown-block(%u,%u) %ld", | |
1107 | MAJOR(dev), MINOR(dev), PTR_ERR(bdev)); | |
ac27a0ec DK |
1108 | return NULL; |
1109 | } | |
1110 | ||
1111 | /* | |
1112 | * Release the journal device | |
1113 | */ | |
4385bab1 | 1114 | static void ext4_blkdev_put(struct block_device *bdev) |
ac27a0ec | 1115 | { |
4385bab1 | 1116 | blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); |
ac27a0ec DK |
1117 | } |
1118 | ||
4385bab1 | 1119 | static void ext4_blkdev_remove(struct ext4_sb_info *sbi) |
ac27a0ec DK |
1120 | { |
1121 | struct block_device *bdev; | |
ee7ed3aa | 1122 | bdev = sbi->s_journal_bdev; |
ac27a0ec | 1123 | if (bdev) { |
4385bab1 | 1124 | ext4_blkdev_put(bdev); |
ee7ed3aa | 1125 | sbi->s_journal_bdev = NULL; |
ac27a0ec | 1126 | } |
ac27a0ec DK |
1127 | } |
1128 | ||
1129 | static inline struct inode *orphan_list_entry(struct list_head *l) | |
1130 | { | |
617ba13b | 1131 | return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; |
ac27a0ec DK |
1132 | } |
1133 | ||
617ba13b | 1134 | static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) |
ac27a0ec DK |
1135 | { |
1136 | struct list_head *l; | |
1137 | ||
b31e1552 ES |
1138 | ext4_msg(sb, KERN_ERR, "sb orphan head is %d", |
1139 | le32_to_cpu(sbi->s_es->s_last_orphan)); | |
ac27a0ec DK |
1140 | |
1141 | printk(KERN_ERR "sb_info orphan list:\n"); | |
1142 | list_for_each(l, &sbi->s_orphan) { | |
1143 | struct inode *inode = orphan_list_entry(l); | |
1144 | printk(KERN_ERR " " | |
1145 | "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", | |
1146 | inode->i_sb->s_id, inode->i_ino, inode, | |
1147 | inode->i_mode, inode->i_nlink, | |
1148 | NEXT_ORPHAN(inode)); | |
1149 | } | |
1150 | } | |
1151 | ||
957153fc JK |
1152 | #ifdef CONFIG_QUOTA |
1153 | static int ext4_quota_off(struct super_block *sb, int type); | |
1154 | ||
1155 | static inline void ext4_quota_off_umount(struct super_block *sb) | |
1156 | { | |
1157 | int type; | |
1158 | ||
964edf66 JK |
1159 | /* Use our quota_off function to clear inode flags etc. */ |
1160 | for (type = 0; type < EXT4_MAXQUOTAS; type++) | |
1161 | ext4_quota_off(sb, type); | |
957153fc | 1162 | } |
33458eab TT |
1163 | |
1164 | /* | |
1165 | * This is a helper function which is used in the mount/remount | |
1166 | * codepaths (which holds s_umount) to fetch the quota file name. | |
1167 | */ | |
1168 | static inline char *get_qf_name(struct super_block *sb, | |
1169 | struct ext4_sb_info *sbi, | |
1170 | int type) | |
1171 | { | |
1172 | return rcu_dereference_protected(sbi->s_qf_names[type], | |
1173 | lockdep_is_held(&sb->s_umount)); | |
1174 | } | |
957153fc JK |
1175 | #else |
1176 | static inline void ext4_quota_off_umount(struct super_block *sb) | |
1177 | { | |
1178 | } | |
1179 | #endif | |
1180 | ||
2b2d6d01 | 1181 | static void ext4_put_super(struct super_block *sb) |
ac27a0ec | 1182 | { |
617ba13b MC |
1183 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
1184 | struct ext4_super_block *es = sbi->s_es; | |
1d0c3924 | 1185 | struct buffer_head **group_desc; |
7c990728 | 1186 | struct flex_groups **flex_groups; |
97abd7d4 | 1187 | int aborted = 0; |
ef2cabf7 | 1188 | int i, err; |
ac27a0ec | 1189 | |
857ac889 | 1190 | ext4_unregister_li_request(sb); |
957153fc | 1191 | ext4_quota_off_umount(sb); |
e0ccfd95 | 1192 | |
c92dc856 | 1193 | flush_work(&sbi->s_error_work); |
2e8fa54e | 1194 | destroy_workqueue(sbi->rsv_conversion_wq); |
02f310fc | 1195 | ext4_release_orphan_info(sb); |
4c0425ff | 1196 | |
5e47868f RH |
1197 | /* |
1198 | * Unregister sysfs before destroying jbd2 journal. | |
1199 | * Since we could still access attr_journal_task attribute via sysfs | |
1200 | * path which could have sbi->s_journal->j_task as NULL | |
1201 | */ | |
1202 | ext4_unregister_sysfs(sb); | |
1203 | ||
0390131b | 1204 | if (sbi->s_journal) { |
97abd7d4 | 1205 | aborted = is_journal_aborted(sbi->s_journal); |
0390131b FM |
1206 | err = jbd2_journal_destroy(sbi->s_journal); |
1207 | sbi->s_journal = NULL; | |
878520ac | 1208 | if ((err < 0) && !aborted) { |
54d3adbc | 1209 | ext4_abort(sb, -err, "Couldn't clean up the journal"); |
878520ac | 1210 | } |
0390131b | 1211 | } |
d4edac31 | 1212 | |
d3922a77 | 1213 | ext4_es_unregister_shrinker(sbi); |
9105bb14 | 1214 | del_timer_sync(&sbi->s_err_report); |
d4edac31 JB |
1215 | ext4_release_system_zone(sb); |
1216 | ext4_mb_release(sb); | |
1217 | ext4_ext_release(sb); | |
d4edac31 | 1218 | |
bc98a42c | 1219 | if (!sb_rdonly(sb) && !aborted) { |
e2b911c5 | 1220 | ext4_clear_feature_journal_needs_recovery(sb); |
02f310fc | 1221 | ext4_clear_feature_orphan_present(sb); |
ac27a0ec | 1222 | es->s_state = cpu_to_le16(sbi->s_mount_state); |
ac27a0ec | 1223 | } |
bc98a42c | 1224 | if (!sb_rdonly(sb)) |
4392fbc4 | 1225 | ext4_commit_super(sb); |
a8e25a83 | 1226 | |
1d0c3924 TT |
1227 | rcu_read_lock(); |
1228 | group_desc = rcu_dereference(sbi->s_group_desc); | |
ac27a0ec | 1229 | for (i = 0; i < sbi->s_gdb_count; i++) |
1d0c3924 TT |
1230 | brelse(group_desc[i]); |
1231 | kvfree(group_desc); | |
7c990728 SJS |
1232 | flex_groups = rcu_dereference(sbi->s_flex_groups); |
1233 | if (flex_groups) { | |
1234 | for (i = 0; i < sbi->s_flex_groups_allocated; i++) | |
1235 | kvfree(flex_groups[i]); | |
1236 | kvfree(flex_groups); | |
1237 | } | |
1d0c3924 | 1238 | rcu_read_unlock(); |
57042651 | 1239 | percpu_counter_destroy(&sbi->s_freeclusters_counter); |
ac27a0ec DK |
1240 | percpu_counter_destroy(&sbi->s_freeinodes_counter); |
1241 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
57042651 | 1242 | percpu_counter_destroy(&sbi->s_dirtyclusters_counter); |
efc61345 | 1243 | percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit); |
bbd55937 | 1244 | percpu_free_rwsem(&sbi->s_writepages_rwsem); |
ac27a0ec | 1245 | #ifdef CONFIG_QUOTA |
a2d4a646 | 1246 | for (i = 0; i < EXT4_MAXQUOTAS; i++) |
33458eab | 1247 | kfree(get_qf_name(sb, sbi, i)); |
ac27a0ec DK |
1248 | #endif |
1249 | ||
1250 | /* Debugging code just in case the in-memory inode orphan list | |
1251 | * isn't empty. The on-disk one can be non-empty if we've | |
1252 | * detected an error and taken the fs readonly, but the | |
1253 | * in-memory list had better be clean by this point. */ | |
1254 | if (!list_empty(&sbi->s_orphan)) | |
1255 | dump_orphan_list(sb, sbi); | |
837c23fb | 1256 | ASSERT(list_empty(&sbi->s_orphan)); |
ac27a0ec | 1257 | |
89d96a6f | 1258 | sync_blockdev(sb->s_bdev); |
f98393a6 | 1259 | invalidate_bdev(sb->s_bdev); |
ee7ed3aa | 1260 | if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) { |
ac27a0ec DK |
1261 | /* |
1262 | * Invalidate the journal device's buffers. We don't want them | |
1263 | * floating about in memory - the physical journal device may | |
1264 | * hotswapped, and it breaks the `ro-after' testing code. | |
1265 | */ | |
ee7ed3aa CX |
1266 | sync_blockdev(sbi->s_journal_bdev); |
1267 | invalidate_bdev(sbi->s_journal_bdev); | |
617ba13b | 1268 | ext4_blkdev_remove(sbi); |
ac27a0ec | 1269 | } |
50c15df6 CX |
1270 | |
1271 | ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); | |
1272 | sbi->s_ea_inode_cache = NULL; | |
1273 | ||
1274 | ext4_xattr_destroy_cache(sbi->s_ea_block_cache); | |
1275 | sbi->s_ea_block_cache = NULL; | |
1276 | ||
618f0031 PS |
1277 | ext4_stop_mmpd(sbi); |
1278 | ||
9060dd2c | 1279 | brelse(sbi->s_sbh); |
ac27a0ec | 1280 | sb->s_fs_info = NULL; |
3197ebdb TT |
1281 | /* |
1282 | * Now that we are completely done shutting down the | |
1283 | * superblock, we need to actually destroy the kobject. | |
1284 | */ | |
3197ebdb TT |
1285 | kobject_put(&sbi->s_kobj); |
1286 | wait_for_completion(&sbi->s_kobj_unregister); | |
0441984a DW |
1287 | if (sbi->s_chksum_driver) |
1288 | crypto_free_shash(sbi->s_chksum_driver); | |
705895b6 | 1289 | kfree(sbi->s_blockgroup_lock); |
5e405595 | 1290 | fs_put_dax(sbi->s_daxdev); |
ac4acb1f | 1291 | fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy); |
c83ad55e | 1292 | #ifdef CONFIG_UNICODE |
f8f4acb6 | 1293 | utf8_unload(sb->s_encoding); |
c83ad55e | 1294 | #endif |
ac27a0ec | 1295 | kfree(sbi); |
ac27a0ec DK |
1296 | } |
1297 | ||
e18b890b | 1298 | static struct kmem_cache *ext4_inode_cachep; |
ac27a0ec DK |
1299 | |
1300 | /* | |
1301 | * Called inside transaction, so use GFP_NOFS | |
1302 | */ | |
617ba13b | 1303 | static struct inode *ext4_alloc_inode(struct super_block *sb) |
ac27a0ec | 1304 | { |
617ba13b | 1305 | struct ext4_inode_info *ei; |
ac27a0ec | 1306 | |
e6b4f8da | 1307 | ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); |
ac27a0ec DK |
1308 | if (!ei) |
1309 | return NULL; | |
0b8e58a1 | 1310 | |
ee73f9a5 | 1311 | inode_set_iversion(&ei->vfs_inode, 1); |
202ee5df | 1312 | spin_lock_init(&ei->i_raw_lock); |
c9de560d | 1313 | INIT_LIST_HEAD(&ei->i_prealloc_list); |
27bc446e | 1314 | atomic_set(&ei->i_prealloc_active, 0); |
c9de560d | 1315 | spin_lock_init(&ei->i_prealloc_lock); |
9a26b661 ZL |
1316 | ext4_es_init_tree(&ei->i_es_tree); |
1317 | rwlock_init(&ei->i_es_lock); | |
edaa53ca | 1318 | INIT_LIST_HEAD(&ei->i_es_list); |
eb68d0e2 | 1319 | ei->i_es_all_nr = 0; |
edaa53ca | 1320 | ei->i_es_shk_nr = 0; |
dd475925 | 1321 | ei->i_es_shrink_lblk = 0; |
d2a17637 | 1322 | ei->i_reserved_data_blocks = 0; |
d2a17637 | 1323 | spin_lock_init(&(ei->i_block_reservation_lock)); |
1dc0aa46 | 1324 | ext4_init_pending_tree(&ei->i_pending_tree); |
a9e7f447 DM |
1325 | #ifdef CONFIG_QUOTA |
1326 | ei->i_reserved_quota = 0; | |
96c7e0d9 | 1327 | memset(&ei->i_dquot, 0, sizeof(ei->i_dquot)); |
a9e7f447 | 1328 | #endif |
8aefcd55 | 1329 | ei->jinode = NULL; |
2e8fa54e | 1330 | INIT_LIST_HEAD(&ei->i_rsv_conversion_list); |
744692dc | 1331 | spin_lock_init(&ei->i_completed_io_lock); |
b436b9be JK |
1332 | ei->i_sync_tid = 0; |
1333 | ei->i_datasync_tid = 0; | |
e27f41e1 | 1334 | atomic_set(&ei->i_unwritten, 0); |
2e8fa54e | 1335 | INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); |
aa75f4d3 HS |
1336 | ext4_fc_init_inode(&ei->vfs_inode); |
1337 | mutex_init(&ei->i_fc_lock); | |
ac27a0ec DK |
1338 | return &ei->vfs_inode; |
1339 | } | |
1340 | ||
7ff9c073 TT |
1341 | static int ext4_drop_inode(struct inode *inode) |
1342 | { | |
1343 | int drop = generic_drop_inode(inode); | |
1344 | ||
29b3692e EB |
1345 | if (!drop) |
1346 | drop = fscrypt_drop_inode(inode); | |
1347 | ||
7ff9c073 TT |
1348 | trace_ext4_drop_inode(inode, drop); |
1349 | return drop; | |
1350 | } | |
1351 | ||
94053139 | 1352 | static void ext4_free_in_core_inode(struct inode *inode) |
fa0d7e3d | 1353 | { |
2c58d548 | 1354 | fscrypt_free_inode(inode); |
aa75f4d3 HS |
1355 | if (!list_empty(&(EXT4_I(inode)->i_fc_list))) { |
1356 | pr_warn("%s: inode %ld still in fc list", | |
1357 | __func__, inode->i_ino); | |
1358 | } | |
fa0d7e3d NP |
1359 | kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); |
1360 | } | |
1361 | ||
617ba13b | 1362 | static void ext4_destroy_inode(struct inode *inode) |
ac27a0ec | 1363 | { |
9f7dd93d | 1364 | if (!list_empty(&(EXT4_I(inode)->i_orphan))) { |
b31e1552 ES |
1365 | ext4_msg(inode->i_sb, KERN_ERR, |
1366 | "Inode %lu (%p): orphan list check failed!", | |
1367 | inode->i_ino, EXT4_I(inode)); | |
9f7dd93d VA |
1368 | print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, |
1369 | EXT4_I(inode), sizeof(struct ext4_inode_info), | |
1370 | true); | |
1371 | dump_stack(); | |
1372 | } | |
6fed8395 JX |
1373 | |
1374 | if (EXT4_I(inode)->i_reserved_data_blocks) | |
1375 | ext4_msg(inode->i_sb, KERN_ERR, | |
1376 | "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!", | |
1377 | inode->i_ino, EXT4_I(inode), | |
1378 | EXT4_I(inode)->i_reserved_data_blocks); | |
ac27a0ec DK |
1379 | } |
1380 | ||
51cc5068 | 1381 | static void init_once(void *foo) |
ac27a0ec | 1382 | { |
617ba13b | 1383 | struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; |
ac27a0ec | 1384 | |
a35afb83 | 1385 | INIT_LIST_HEAD(&ei->i_orphan); |
a35afb83 | 1386 | init_rwsem(&ei->xattr_sem); |
0e855ac8 | 1387 | init_rwsem(&ei->i_data_sem); |
a35afb83 | 1388 | inode_init_once(&ei->vfs_inode); |
aa75f4d3 | 1389 | ext4_fc_init_inode(&ei->vfs_inode); |
ac27a0ec DK |
1390 | } |
1391 | ||
e67bc2b3 | 1392 | static int __init init_inodecache(void) |
ac27a0ec | 1393 | { |
f8dd7c70 DW |
1394 | ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache", |
1395 | sizeof(struct ext4_inode_info), 0, | |
1396 | (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD| | |
1397 | SLAB_ACCOUNT), | |
1398 | offsetof(struct ext4_inode_info, i_data), | |
1399 | sizeof_field(struct ext4_inode_info, i_data), | |
1400 | init_once); | |
617ba13b | 1401 | if (ext4_inode_cachep == NULL) |
ac27a0ec DK |
1402 | return -ENOMEM; |
1403 | return 0; | |
1404 | } | |
1405 | ||
1406 | static void destroy_inodecache(void) | |
1407 | { | |
8c0a8537 KS |
1408 | /* |
1409 | * Make sure all delayed rcu free inodes are flushed before we | |
1410 | * destroy cache. | |
1411 | */ | |
1412 | rcu_barrier(); | |
617ba13b | 1413 | kmem_cache_destroy(ext4_inode_cachep); |
ac27a0ec DK |
1414 | } |
1415 | ||
0930fcc1 | 1416 | void ext4_clear_inode(struct inode *inode) |
ac27a0ec | 1417 | { |
aa75f4d3 | 1418 | ext4_fc_del(inode); |
0930fcc1 | 1419 | invalidate_inode_buffers(inode); |
dbd5768f | 1420 | clear_inode(inode); |
27bc446e | 1421 | ext4_discard_preallocations(inode, 0); |
51865fda | 1422 | ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); |
f4c2d372 | 1423 | dquot_drop(inode); |
8aefcd55 TT |
1424 | if (EXT4_I(inode)->jinode) { |
1425 | jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), | |
1426 | EXT4_I(inode)->jinode); | |
1427 | jbd2_free_inode(EXT4_I(inode)->jinode); | |
1428 | EXT4_I(inode)->jinode = NULL; | |
1429 | } | |
3d204e24 | 1430 | fscrypt_put_encryption_info(inode); |
c93d8f88 | 1431 | fsverity_cleanup_inode(inode); |
ac27a0ec DK |
1432 | } |
1433 | ||
1b961ac0 | 1434 | static struct inode *ext4_nfs_get_inode(struct super_block *sb, |
0b8e58a1 | 1435 | u64 ino, u32 generation) |
ac27a0ec | 1436 | { |
ac27a0ec | 1437 | struct inode *inode; |
ac27a0ec | 1438 | |
8a363970 | 1439 | /* |
ac27a0ec DK |
1440 | * Currently we don't know the generation for parent directory, so |
1441 | * a generation of 0 means "accept any" | |
1442 | */ | |
8a363970 | 1443 | inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE); |
1d1fe1ee DH |
1444 | if (IS_ERR(inode)) |
1445 | return ERR_CAST(inode); | |
1446 | if (generation && inode->i_generation != generation) { | |
ac27a0ec DK |
1447 | iput(inode); |
1448 | return ERR_PTR(-ESTALE); | |
1449 | } | |
1b961ac0 CH |
1450 | |
1451 | return inode; | |
1452 | } | |
1453 | ||
1454 | static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, | |
0b8e58a1 | 1455 | int fh_len, int fh_type) |
1b961ac0 CH |
1456 | { |
1457 | return generic_fh_to_dentry(sb, fid, fh_len, fh_type, | |
1458 | ext4_nfs_get_inode); | |
1459 | } | |
1460 | ||
1461 | static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, | |
0b8e58a1 | 1462 | int fh_len, int fh_type) |
1b961ac0 CH |
1463 | { |
1464 | return generic_fh_to_parent(sb, fid, fh_len, fh_type, | |
1465 | ext4_nfs_get_inode); | |
ac27a0ec DK |
1466 | } |
1467 | ||
fde87268 TT |
1468 | static int ext4_nfs_commit_metadata(struct inode *inode) |
1469 | { | |
1470 | struct writeback_control wbc = { | |
1471 | .sync_mode = WB_SYNC_ALL | |
1472 | }; | |
1473 | ||
1474 | trace_ext4_nfs_commit_metadata(inode); | |
1475 | return ext4_write_inode(inode, &wbc); | |
1476 | } | |
1477 | ||
643fa961 | 1478 | #ifdef CONFIG_FS_ENCRYPTION |
a7550b30 JK |
1479 | static int ext4_get_context(struct inode *inode, void *ctx, size_t len) |
1480 | { | |
1481 | return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION, | |
1482 | EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len); | |
1483 | } | |
1484 | ||
a7550b30 JK |
1485 | static int ext4_set_context(struct inode *inode, const void *ctx, size_t len, |
1486 | void *fs_data) | |
1487 | { | |
2f8f5e76 | 1488 | handle_t *handle = fs_data; |
c1a5d5f6 | 1489 | int res, res2, credits, retries = 0; |
2f8f5e76 | 1490 | |
9ce0151a EB |
1491 | /* |
1492 | * Encrypting the root directory is not allowed because e2fsck expects | |
1493 | * lost+found to exist and be unencrypted, and encrypting the root | |
1494 | * directory would imply encrypting the lost+found directory as well as | |
1495 | * the filename "lost+found" itself. | |
1496 | */ | |
1497 | if (inode->i_ino == EXT4_ROOT_INO) | |
1498 | return -EPERM; | |
2f8f5e76 | 1499 | |
7d3e06a8 RZ |
1500 | if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode))) |
1501 | return -EINVAL; | |
1502 | ||
b383a73f IW |
1503 | if (ext4_test_inode_flag(inode, EXT4_INODE_DAX)) |
1504 | return -EOPNOTSUPP; | |
1505 | ||
94840e3c EB |
1506 | res = ext4_convert_inline_data(inode); |
1507 | if (res) | |
1508 | return res; | |
1509 | ||
2f8f5e76 EB |
1510 | /* |
1511 | * If a journal handle was specified, then the encryption context is | |
1512 | * being set on a new inode via inheritance and is part of a larger | |
1513 | * transaction to create the inode. Otherwise the encryption context is | |
1514 | * being set on an existing inode in its own transaction. Only in the | |
1515 | * latter case should the "retry on ENOSPC" logic be used. | |
1516 | */ | |
a7550b30 | 1517 | |
2f8f5e76 EB |
1518 | if (handle) { |
1519 | res = ext4_xattr_set_handle(handle, inode, | |
1520 | EXT4_XATTR_INDEX_ENCRYPTION, | |
1521 | EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, | |
1522 | ctx, len, 0); | |
a7550b30 JK |
1523 | if (!res) { |
1524 | ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); | |
1525 | ext4_clear_inode_state(inode, | |
1526 | EXT4_STATE_MAY_INLINE_DATA); | |
a3caa24b | 1527 | /* |
2ee6a576 EB |
1528 | * Update inode->i_flags - S_ENCRYPTED will be enabled, |
1529 | * S_DAX may be disabled | |
a3caa24b | 1530 | */ |
043546e4 | 1531 | ext4_set_inode_flags(inode, false); |
a7550b30 JK |
1532 | } |
1533 | return res; | |
1534 | } | |
1535 | ||
b8cb5a54 TE |
1536 | res = dquot_initialize(inode); |
1537 | if (res) | |
1538 | return res; | |
2f8f5e76 | 1539 | retry: |
af65207c TE |
1540 | res = ext4_xattr_set_credits(inode, len, false /* is_create */, |
1541 | &credits); | |
dec214d0 TE |
1542 | if (res) |
1543 | return res; | |
1544 | ||
c1a5d5f6 | 1545 | handle = ext4_journal_start(inode, EXT4_HT_MISC, credits); |
a7550b30 JK |
1546 | if (IS_ERR(handle)) |
1547 | return PTR_ERR(handle); | |
1548 | ||
2f8f5e76 EB |
1549 | res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION, |
1550 | EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, | |
1551 | ctx, len, 0); | |
a7550b30 JK |
1552 | if (!res) { |
1553 | ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); | |
2ee6a576 EB |
1554 | /* |
1555 | * Update inode->i_flags - S_ENCRYPTED will be enabled, | |
1556 | * S_DAX may be disabled | |
1557 | */ | |
043546e4 | 1558 | ext4_set_inode_flags(inode, false); |
a7550b30 JK |
1559 | res = ext4_mark_inode_dirty(handle, inode); |
1560 | if (res) | |
1561 | EXT4_ERROR_INODE(inode, "Failed to mark inode dirty"); | |
1562 | } | |
1563 | res2 = ext4_journal_stop(handle); | |
2f8f5e76 EB |
1564 | |
1565 | if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) | |
1566 | goto retry; | |
a7550b30 JK |
1567 | if (!res) |
1568 | res = res2; | |
1569 | return res; | |
1570 | } | |
1571 | ||
ac4acb1f | 1572 | static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb) |
a7550b30 | 1573 | { |
ac4acb1f | 1574 | return EXT4_SB(sb)->s_dummy_enc_policy.policy; |
a7550b30 JK |
1575 | } |
1576 | ||
b925acb8 EB |
1577 | static bool ext4_has_stable_inodes(struct super_block *sb) |
1578 | { | |
1579 | return ext4_has_feature_stable_inodes(sb); | |
1580 | } | |
1581 | ||
1582 | static void ext4_get_ino_and_lblk_bits(struct super_block *sb, | |
1583 | int *ino_bits_ret, int *lblk_bits_ret) | |
1584 | { | |
1585 | *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count); | |
1586 | *lblk_bits_ret = 8 * sizeof(ext4_lblk_t); | |
1587 | } | |
1588 | ||
6f69f0ed | 1589 | static const struct fscrypt_operations ext4_cryptops = { |
a5d431ef | 1590 | .key_prefix = "ext4:", |
a7550b30 | 1591 | .get_context = ext4_get_context, |
a7550b30 | 1592 | .set_context = ext4_set_context, |
ac4acb1f | 1593 | .get_dummy_policy = ext4_get_dummy_policy, |
a7550b30 | 1594 | .empty_dir = ext4_empty_dir, |
b925acb8 EB |
1595 | .has_stable_inodes = ext4_has_stable_inodes, |
1596 | .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits, | |
a7550b30 | 1597 | }; |
a7550b30 JK |
1598 | #endif |
1599 | ||
ac27a0ec | 1600 | #ifdef CONFIG_QUOTA |
d6006186 | 1601 | static const char * const quotatypes[] = INITQFNAMES; |
689c958c | 1602 | #define QTYPE2NAME(t) (quotatypes[t]) |
ac27a0ec | 1603 | |
617ba13b MC |
1604 | static int ext4_write_dquot(struct dquot *dquot); |
1605 | static int ext4_acquire_dquot(struct dquot *dquot); | |
1606 | static int ext4_release_dquot(struct dquot *dquot); | |
1607 | static int ext4_mark_dquot_dirty(struct dquot *dquot); | |
1608 | static int ext4_write_info(struct super_block *sb, int type); | |
6f28e087 | 1609 | static int ext4_quota_on(struct super_block *sb, int type, int format_id, |
8c54ca9c | 1610 | const struct path *path); |
617ba13b | 1611 | static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, |
ac27a0ec | 1612 | size_t len, loff_t off); |
617ba13b | 1613 | static ssize_t ext4_quota_write(struct super_block *sb, int type, |
ac27a0ec | 1614 | const char *data, size_t len, loff_t off); |
7c319d32 AK |
1615 | static int ext4_quota_enable(struct super_block *sb, int type, int format_id, |
1616 | unsigned int flags); | |
ac27a0ec | 1617 | |
96c7e0d9 JK |
1618 | static struct dquot **ext4_get_dquots(struct inode *inode) |
1619 | { | |
1620 | return EXT4_I(inode)->i_dquot; | |
1621 | } | |
1622 | ||
61e225dc | 1623 | static const struct dquot_operations ext4_quota_operations = { |
7a9ca53a TE |
1624 | .get_reserved_space = ext4_get_reserved_space, |
1625 | .write_dquot = ext4_write_dquot, | |
1626 | .acquire_dquot = ext4_acquire_dquot, | |
1627 | .release_dquot = ext4_release_dquot, | |
1628 | .mark_dirty = ext4_mark_dquot_dirty, | |
1629 | .write_info = ext4_write_info, | |
1630 | .alloc_dquot = dquot_alloc, | |
1631 | .destroy_dquot = dquot_destroy, | |
1632 | .get_projid = ext4_get_projid, | |
1633 | .get_inode_usage = ext4_get_inode_usage, | |
ebc11f7b | 1634 | .get_next_id = dquot_get_next_id, |
ac27a0ec DK |
1635 | }; |
1636 | ||
0d54b217 | 1637 | static const struct quotactl_ops ext4_qctl_operations = { |
617ba13b | 1638 | .quota_on = ext4_quota_on, |
ca0e05e4 | 1639 | .quota_off = ext4_quota_off, |
287a8095 | 1640 | .quota_sync = dquot_quota_sync, |
0a240339 | 1641 | .get_state = dquot_get_state, |
287a8095 CH |
1642 | .set_info = dquot_set_dqinfo, |
1643 | .get_dqblk = dquot_get_dqblk, | |
6332b9b5 ES |
1644 | .set_dqblk = dquot_set_dqblk, |
1645 | .get_nextdqblk = dquot_get_next_dqblk, | |
ac27a0ec DK |
1646 | }; |
1647 | #endif | |
1648 | ||
ee9b6d61 | 1649 | static const struct super_operations ext4_sops = { |
617ba13b | 1650 | .alloc_inode = ext4_alloc_inode, |
94053139 | 1651 | .free_inode = ext4_free_in_core_inode, |
617ba13b | 1652 | .destroy_inode = ext4_destroy_inode, |
617ba13b MC |
1653 | .write_inode = ext4_write_inode, |
1654 | .dirty_inode = ext4_dirty_inode, | |
7ff9c073 | 1655 | .drop_inode = ext4_drop_inode, |
0930fcc1 | 1656 | .evict_inode = ext4_evict_inode, |
617ba13b | 1657 | .put_super = ext4_put_super, |
617ba13b | 1658 | .sync_fs = ext4_sync_fs, |
c4be0c1d TS |
1659 | .freeze_fs = ext4_freeze, |
1660 | .unfreeze_fs = ext4_unfreeze, | |
617ba13b MC |
1661 | .statfs = ext4_statfs, |
1662 | .remount_fs = ext4_remount, | |
617ba13b | 1663 | .show_options = ext4_show_options, |
ac27a0ec | 1664 | #ifdef CONFIG_QUOTA |
617ba13b MC |
1665 | .quota_read = ext4_quota_read, |
1666 | .quota_write = ext4_quota_write, | |
96c7e0d9 | 1667 | .get_dquots = ext4_get_dquots, |
ac27a0ec DK |
1668 | #endif |
1669 | }; | |
1670 | ||
39655164 | 1671 | static const struct export_operations ext4_export_ops = { |
1b961ac0 CH |
1672 | .fh_to_dentry = ext4_fh_to_dentry, |
1673 | .fh_to_parent = ext4_fh_to_parent, | |
617ba13b | 1674 | .get_parent = ext4_get_parent, |
fde87268 | 1675 | .commit_metadata = ext4_nfs_commit_metadata, |
ac27a0ec DK |
1676 | }; |
1677 | ||
1678 | enum { | |
1679 | Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, | |
1680 | Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, | |
72578c33 | 1681 | Opt_nouid32, Opt_debug, Opt_removed, |
ac27a0ec | 1682 | Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, |
72578c33 | 1683 | Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, |
ad4eec61 ES |
1684 | Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, |
1685 | Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, | |
ac27a0ec | 1686 | Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, |
6ddb2447 | 1687 | Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, |
4f74d15f | 1688 | Opt_inlinecrypt, |
ac27a0ec | 1689 | Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, |
5a20bdfc | 1690 | Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, |
ee4a3fcd | 1691 | Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, |
9cb20f94 IW |
1692 | Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, |
1693 | Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never, | |
327eaf73 TT |
1694 | Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error, |
1695 | Opt_nowarn_on_error, Opt_mblk_io_submit, | |
670e9875 | 1696 | Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize, |
1449032b | 1697 | Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, |
5328e635 | 1698 | Opt_inode_readahead_blks, Opt_journal_ioprio, |
744692dc | 1699 | Opt_dioread_nolock, Opt_dioread_lock, |
fc6cb1cd | 1700 | Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, |
cdb7ee4c | 1701 | Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache, |
21175ca4 | 1702 | Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan, |
e5a185c2 | 1703 | Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type, |
8016e29f | 1704 | #ifdef CONFIG_EXT4_DEBUG |
99c880de | 1705 | Opt_fc_debug_max_replay, Opt_fc_debug_force |
8016e29f | 1706 | #endif |
ac27a0ec DK |
1707 | }; |
1708 | ||
e5a185c2 LC |
1709 | static const struct constant_table ext4_param_errors[] = { |
1710 | {"continue", Opt_err_cont}, | |
1711 | {"panic", Opt_err_panic}, | |
1712 | {"remount-ro", Opt_err_ro}, | |
1713 | {} | |
1714 | }; | |
1715 | ||
1716 | static const struct constant_table ext4_param_data[] = { | |
1717 | {"journal", Opt_data_journal}, | |
1718 | {"ordered", Opt_data_ordered}, | |
1719 | {"writeback", Opt_data_writeback}, | |
1720 | {} | |
1721 | }; | |
1722 | ||
1723 | static const struct constant_table ext4_param_data_err[] = { | |
1724 | {"abort", Opt_data_err_abort}, | |
1725 | {"ignore", Opt_data_err_ignore}, | |
1726 | {} | |
1727 | }; | |
1728 | ||
1729 | static const struct constant_table ext4_param_jqfmt[] = { | |
1730 | {"vfsold", Opt_jqfmt_vfsold}, | |
1731 | {"vfsv0", Opt_jqfmt_vfsv0}, | |
1732 | {"vfsv1", Opt_jqfmt_vfsv1}, | |
1733 | {} | |
1734 | }; | |
1735 | ||
1736 | static const struct constant_table ext4_param_dax[] = { | |
1737 | {"always", Opt_dax_always}, | |
1738 | {"inode", Opt_dax_inode}, | |
1739 | {"never", Opt_dax_never}, | |
1740 | {} | |
1741 | }; | |
1742 | ||
1743 | /* String parameter that allows empty argument */ | |
1744 | #define fsparam_string_empty(NAME, OPT) \ | |
1745 | __fsparam(fs_param_is_string, NAME, OPT, fs_param_can_be_empty, NULL) | |
1746 | ||
1747 | /* | |
1748 | * Mount option specification | |
1749 | * We don't use fsparam_flag_no because of the way we set the | |
1750 | * options and the way we show them in _ext4_show_options(). To | |
1751 | * keep the changes to a minimum, let's keep the negative options | |
1752 | * separate for now. | |
1753 | */ | |
1754 | static const struct fs_parameter_spec ext4_param_specs[] = { | |
1755 | fsparam_flag ("bsddf", Opt_bsd_df), | |
1756 | fsparam_flag ("minixdf", Opt_minix_df), | |
1757 | fsparam_flag ("grpid", Opt_grpid), | |
1758 | fsparam_flag ("bsdgroups", Opt_grpid), | |
1759 | fsparam_flag ("nogrpid", Opt_nogrpid), | |
1760 | fsparam_flag ("sysvgroups", Opt_nogrpid), | |
1761 | fsparam_u32 ("resgid", Opt_resgid), | |
1762 | fsparam_u32 ("resuid", Opt_resuid), | |
1763 | fsparam_u32 ("sb", Opt_sb), | |
1764 | fsparam_enum ("errors", Opt_errors, ext4_param_errors), | |
1765 | fsparam_flag ("nouid32", Opt_nouid32), | |
1766 | fsparam_flag ("debug", Opt_debug), | |
1767 | fsparam_flag ("oldalloc", Opt_removed), | |
1768 | fsparam_flag ("orlov", Opt_removed), | |
1769 | fsparam_flag ("user_xattr", Opt_user_xattr), | |
1770 | fsparam_flag ("nouser_xattr", Opt_nouser_xattr), | |
1771 | fsparam_flag ("acl", Opt_acl), | |
1772 | fsparam_flag ("noacl", Opt_noacl), | |
1773 | fsparam_flag ("norecovery", Opt_noload), | |
1774 | fsparam_flag ("noload", Opt_noload), | |
1775 | fsparam_flag ("bh", Opt_removed), | |
1776 | fsparam_flag ("nobh", Opt_removed), | |
1777 | fsparam_u32 ("commit", Opt_commit), | |
1778 | fsparam_u32 ("min_batch_time", Opt_min_batch_time), | |
1779 | fsparam_u32 ("max_batch_time", Opt_max_batch_time), | |
1780 | fsparam_u32 ("journal_dev", Opt_journal_dev), | |
1781 | fsparam_bdev ("journal_path", Opt_journal_path), | |
1782 | fsparam_flag ("journal_checksum", Opt_journal_checksum), | |
1783 | fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum), | |
1784 | fsparam_flag ("journal_async_commit",Opt_journal_async_commit), | |
1785 | fsparam_flag ("abort", Opt_abort), | |
1786 | fsparam_enum ("data", Opt_data, ext4_param_data), | |
1787 | fsparam_enum ("data_err", Opt_data_err, | |
1788 | ext4_param_data_err), | |
1789 | fsparam_string_empty | |
1790 | ("usrjquota", Opt_usrjquota), | |
1791 | fsparam_string_empty | |
1792 | ("grpjquota", Opt_grpjquota), | |
1793 | fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt), | |
1794 | fsparam_flag ("grpquota", Opt_grpquota), | |
1795 | fsparam_flag ("quota", Opt_quota), | |
1796 | fsparam_flag ("noquota", Opt_noquota), | |
1797 | fsparam_flag ("usrquota", Opt_usrquota), | |
1798 | fsparam_flag ("prjquota", Opt_prjquota), | |
1799 | fsparam_flag ("barrier", Opt_barrier), | |
1800 | fsparam_u32 ("barrier", Opt_barrier), | |
1801 | fsparam_flag ("nobarrier", Opt_nobarrier), | |
1802 | fsparam_flag ("i_version", Opt_i_version), | |
1803 | fsparam_flag ("dax", Opt_dax), | |
1804 | fsparam_enum ("dax", Opt_dax_type, ext4_param_dax), | |
1805 | fsparam_u32 ("stripe", Opt_stripe), | |
1806 | fsparam_flag ("delalloc", Opt_delalloc), | |
1807 | fsparam_flag ("nodelalloc", Opt_nodelalloc), | |
1808 | fsparam_flag ("warn_on_error", Opt_warn_on_error), | |
1809 | fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error), | |
1810 | fsparam_flag ("lazytime", Opt_lazytime), | |
1811 | fsparam_flag ("nolazytime", Opt_nolazytime), | |
1812 | fsparam_u32 ("debug_want_extra_isize", | |
1813 | Opt_debug_want_extra_isize), | |
1814 | fsparam_flag ("mblk_io_submit", Opt_removed), | |
1815 | fsparam_flag ("nomblk_io_submit", Opt_removed), | |
1816 | fsparam_flag ("block_validity", Opt_block_validity), | |
1817 | fsparam_flag ("noblock_validity", Opt_noblock_validity), | |
1818 | fsparam_u32 ("inode_readahead_blks", | |
1819 | Opt_inode_readahead_blks), | |
1820 | fsparam_u32 ("journal_ioprio", Opt_journal_ioprio), | |
1821 | fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc), | |
1822 | fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc), | |
1823 | fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc), | |
1824 | fsparam_flag ("dioread_nolock", Opt_dioread_nolock), | |
1825 | fsparam_flag ("nodioread_nolock", Opt_dioread_lock), | |
1826 | fsparam_flag ("dioread_lock", Opt_dioread_lock), | |
1827 | fsparam_flag ("discard", Opt_discard), | |
1828 | fsparam_flag ("nodiscard", Opt_nodiscard), | |
1829 | fsparam_u32 ("init_itable", Opt_init_itable), | |
1830 | fsparam_flag ("init_itable", Opt_init_itable), | |
1831 | fsparam_flag ("noinit_itable", Opt_noinit_itable), | |
1832 | #ifdef CONFIG_EXT4_DEBUG | |
1833 | fsparam_flag ("fc_debug_force", Opt_fc_debug_force), | |
1834 | fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay), | |
1835 | #endif | |
1836 | fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb), | |
1837 | fsparam_flag ("test_dummy_encryption", | |
1838 | Opt_test_dummy_encryption), | |
1839 | fsparam_string ("test_dummy_encryption", | |
1840 | Opt_test_dummy_encryption), | |
1841 | fsparam_flag ("inlinecrypt", Opt_inlinecrypt), | |
1842 | fsparam_flag ("nombcache", Opt_nombcache), | |
1843 | fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */ | |
1844 | fsparam_flag ("prefetch_block_bitmaps", | |
1845 | Opt_removed), | |
1846 | fsparam_flag ("no_prefetch_block_bitmaps", | |
1847 | Opt_no_prefetch_block_bitmaps), | |
1848 | fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan), | |
1849 | fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */ | |
1850 | fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */ | |
1851 | fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */ | |
1852 | fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */ | |
1853 | fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */ | |
1854 | {} | |
1855 | }; | |
1856 | ||
a447c093 | 1857 | static const match_table_t tokens = { |
ac27a0ec DK |
1858 | {Opt_bsd_df, "bsddf"}, |
1859 | {Opt_minix_df, "minixdf"}, | |
1860 | {Opt_grpid, "grpid"}, | |
1861 | {Opt_grpid, "bsdgroups"}, | |
1862 | {Opt_nogrpid, "nogrpid"}, | |
1863 | {Opt_nogrpid, "sysvgroups"}, | |
1864 | {Opt_resgid, "resgid=%u"}, | |
1865 | {Opt_resuid, "resuid=%u"}, | |
1866 | {Opt_sb, "sb=%u"}, | |
1867 | {Opt_err_cont, "errors=continue"}, | |
1868 | {Opt_err_panic, "errors=panic"}, | |
1869 | {Opt_err_ro, "errors=remount-ro"}, | |
1870 | {Opt_nouid32, "nouid32"}, | |
ac27a0ec | 1871 | {Opt_debug, "debug"}, |
72578c33 TT |
1872 | {Opt_removed, "oldalloc"}, |
1873 | {Opt_removed, "orlov"}, | |
ac27a0ec DK |
1874 | {Opt_user_xattr, "user_xattr"}, |
1875 | {Opt_nouser_xattr, "nouser_xattr"}, | |
1876 | {Opt_acl, "acl"}, | |
1877 | {Opt_noacl, "noacl"}, | |
e3bb52ae | 1878 | {Opt_noload, "norecovery"}, |
5a916be1 | 1879 | {Opt_noload, "noload"}, |
72578c33 TT |
1880 | {Opt_removed, "nobh"}, |
1881 | {Opt_removed, "bh"}, | |
ac27a0ec | 1882 | {Opt_commit, "commit=%u"}, |
30773840 TT |
1883 | {Opt_min_batch_time, "min_batch_time=%u"}, |
1884 | {Opt_max_batch_time, "max_batch_time=%u"}, | |
ac27a0ec | 1885 | {Opt_journal_dev, "journal_dev=%u"}, |
ad4eec61 | 1886 | {Opt_journal_path, "journal_path=%s"}, |
818d276c | 1887 | {Opt_journal_checksum, "journal_checksum"}, |
c6d3d56d | 1888 | {Opt_nojournal_checksum, "nojournal_checksum"}, |
818d276c | 1889 | {Opt_journal_async_commit, "journal_async_commit"}, |
ac27a0ec DK |
1890 | {Opt_abort, "abort"}, |
1891 | {Opt_data_journal, "data=journal"}, | |
1892 | {Opt_data_ordered, "data=ordered"}, | |
1893 | {Opt_data_writeback, "data=writeback"}, | |
5bf5683a HK |
1894 | {Opt_data_err_abort, "data_err=abort"}, |
1895 | {Opt_data_err_ignore, "data_err=ignore"}, | |
ac27a0ec DK |
1896 | {Opt_offusrjquota, "usrjquota="}, |
1897 | {Opt_usrjquota, "usrjquota=%s"}, | |
1898 | {Opt_offgrpjquota, "grpjquota="}, | |
1899 | {Opt_grpjquota, "grpjquota=%s"}, | |
1900 | {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, | |
1901 | {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, | |
5a20bdfc | 1902 | {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, |
ac27a0ec DK |
1903 | {Opt_grpquota, "grpquota"}, |
1904 | {Opt_noquota, "noquota"}, | |
1905 | {Opt_quota, "quota"}, | |
1906 | {Opt_usrquota, "usrquota"}, | |
49da9392 | 1907 | {Opt_prjquota, "prjquota"}, |
ac27a0ec | 1908 | {Opt_barrier, "barrier=%u"}, |
06705bff TT |
1909 | {Opt_barrier, "barrier"}, |
1910 | {Opt_nobarrier, "nobarrier"}, | |
25ec56b5 | 1911 | {Opt_i_version, "i_version"}, |
923ae0ff | 1912 | {Opt_dax, "dax"}, |
9cb20f94 IW |
1913 | {Opt_dax_always, "dax=always"}, |
1914 | {Opt_dax_inode, "dax=inode"}, | |
1915 | {Opt_dax_never, "dax=never"}, | |
c9de560d | 1916 | {Opt_stripe, "stripe=%u"}, |
64769240 | 1917 | {Opt_delalloc, "delalloc"}, |
327eaf73 TT |
1918 | {Opt_warn_on_error, "warn_on_error"}, |
1919 | {Opt_nowarn_on_error, "nowarn_on_error"}, | |
a26f4992 TT |
1920 | {Opt_lazytime, "lazytime"}, |
1921 | {Opt_nolazytime, "nolazytime"}, | |
670e9875 | 1922 | {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"}, |
dd919b98 | 1923 | {Opt_nodelalloc, "nodelalloc"}, |
36ade451 JK |
1924 | {Opt_removed, "mblk_io_submit"}, |
1925 | {Opt_removed, "nomblk_io_submit"}, | |
6fd058f7 TT |
1926 | {Opt_block_validity, "block_validity"}, |
1927 | {Opt_noblock_validity, "noblock_validity"}, | |
240799cd | 1928 | {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, |
b3881f74 | 1929 | {Opt_journal_ioprio, "journal_ioprio=%u"}, |
afd4672d | 1930 | {Opt_auto_da_alloc, "auto_da_alloc=%u"}, |
06705bff TT |
1931 | {Opt_auto_da_alloc, "auto_da_alloc"}, |
1932 | {Opt_noauto_da_alloc, "noauto_da_alloc"}, | |
744692dc | 1933 | {Opt_dioread_nolock, "dioread_nolock"}, |
244adf64 | 1934 | {Opt_dioread_lock, "nodioread_nolock"}, |
744692dc | 1935 | {Opt_dioread_lock, "dioread_lock"}, |
5328e635 ES |
1936 | {Opt_discard, "discard"}, |
1937 | {Opt_nodiscard, "nodiscard"}, | |
fc6cb1cd TT |
1938 | {Opt_init_itable, "init_itable=%u"}, |
1939 | {Opt_init_itable, "init_itable"}, | |
1940 | {Opt_noinit_itable, "noinit_itable"}, | |
8016e29f | 1941 | #ifdef CONFIG_EXT4_DEBUG |
99c880de | 1942 | {Opt_fc_debug_force, "fc_debug_force"}, |
8016e29f HS |
1943 | {Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"}, |
1944 | #endif | |
df981d03 | 1945 | {Opt_max_dir_size_kb, "max_dir_size_kb=%u"}, |
ed318a6c | 1946 | {Opt_test_dummy_encryption, "test_dummy_encryption=%s"}, |
6ddb2447 | 1947 | {Opt_test_dummy_encryption, "test_dummy_encryption"}, |
4f74d15f | 1948 | {Opt_inlinecrypt, "inlinecrypt"}, |
cdb7ee4c TE |
1949 | {Opt_nombcache, "nombcache"}, |
1950 | {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */ | |
21175ca4 HS |
1951 | {Opt_removed, "prefetch_block_bitmaps"}, |
1952 | {Opt_no_prefetch_block_bitmaps, "no_prefetch_block_bitmaps"}, | |
196e402a | 1953 | {Opt_mb_optimize_scan, "mb_optimize_scan=%d"}, |
c7198b9c TT |
1954 | {Opt_removed, "check=none"}, /* mount option from ext2/3 */ |
1955 | {Opt_removed, "nocheck"}, /* mount option from ext2/3 */ | |
1956 | {Opt_removed, "reservation"}, /* mount option from ext2/3 */ | |
1957 | {Opt_removed, "noreservation"}, /* mount option from ext2/3 */ | |
1958 | {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */ | |
f3f12faa | 1959 | {Opt_err, NULL}, |
ac27a0ec DK |
1960 | }; |
1961 | ||
617ba13b | 1962 | static ext4_fsblk_t get_sb_block(void **data) |
ac27a0ec | 1963 | { |
617ba13b | 1964 | ext4_fsblk_t sb_block; |
ac27a0ec DK |
1965 | char *options = (char *) *data; |
1966 | ||
1967 | if (!options || strncmp(options, "sb=", 3) != 0) | |
1968 | return 1; /* Default location */ | |
0b8e58a1 | 1969 | |
ac27a0ec | 1970 | options += 3; |
0b8e58a1 | 1971 | /* TODO: use simple_strtoll with >32bit ext4 */ |
ac27a0ec DK |
1972 | sb_block = simple_strtoul(options, &options, 0); |
1973 | if (*options && *options != ',') { | |
4776004f | 1974 | printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", |
ac27a0ec DK |
1975 | (char *) *data); |
1976 | return 1; | |
1977 | } | |
1978 | if (*options == ',') | |
1979 | options++; | |
1980 | *data = (void *) options; | |
0b8e58a1 | 1981 | |
ac27a0ec DK |
1982 | return sb_block; |
1983 | } | |
1984 | ||
b3881f74 | 1985 | #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) |
196e402a HS |
1986 | #define DEFAULT_MB_OPTIMIZE_SCAN (-1) |
1987 | ||
d6006186 EB |
1988 | static const char deprecated_msg[] = |
1989 | "Mount option \"%s\" will be removed by %s\n" | |
437ca0fd | 1990 | "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; |
b3881f74 | 1991 | |
26092bf5 TT |
1992 | #define MOPT_SET 0x0001 |
1993 | #define MOPT_CLEAR 0x0002 | |
1994 | #define MOPT_NOSUPPORT 0x0004 | |
1995 | #define MOPT_EXPLICIT 0x0008 | |
1996 | #define MOPT_CLEAR_ERR 0x0010 | |
1997 | #define MOPT_GTE0 0x0020 | |
ac27a0ec | 1998 | #ifdef CONFIG_QUOTA |
26092bf5 TT |
1999 | #define MOPT_Q 0 |
2000 | #define MOPT_QFMT 0x0040 | |
2001 | #else | |
2002 | #define MOPT_Q MOPT_NOSUPPORT | |
2003 | #define MOPT_QFMT MOPT_NOSUPPORT | |
ac27a0ec | 2004 | #endif |
26092bf5 | 2005 | #define MOPT_DATAJ 0x0080 |
8dc0aa8c TT |
2006 | #define MOPT_NO_EXT2 0x0100 |
2007 | #define MOPT_NO_EXT3 0x0200 | |
2008 | #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) | |
ad4eec61 | 2009 | #define MOPT_STRING 0x0400 |
9cb20f94 | 2010 | #define MOPT_SKIP 0x0800 |
995a3ed6 | 2011 | #define MOPT_2 0x1000 |
26092bf5 TT |
2012 | |
2013 | static const struct mount_opts { | |
2014 | int token; | |
2015 | int mount_opt; | |
2016 | int flags; | |
2017 | } ext4_mount_opts[] = { | |
2018 | {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, | |
2019 | {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, | |
2020 | {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, | |
2021 | {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, | |
26092bf5 TT |
2022 | {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, |
2023 | {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, | |
8dc0aa8c TT |
2024 | {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, |
2025 | MOPT_EXT4_ONLY | MOPT_SET}, | |
2026 | {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, | |
2027 | MOPT_EXT4_ONLY | MOPT_CLEAR}, | |
26092bf5 TT |
2028 | {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, |
2029 | {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, | |
8dc0aa8c TT |
2030 | {Opt_delalloc, EXT4_MOUNT_DELALLOC, |
2031 | MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, | |
2032 | {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, | |
59d9fa5c | 2033 | MOPT_EXT4_ONLY | MOPT_CLEAR}, |
327eaf73 TT |
2034 | {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET}, |
2035 | {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR}, | |
c6d3d56d DW |
2036 | {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, |
2037 | MOPT_EXT4_ONLY | MOPT_CLEAR}, | |
8dc0aa8c | 2038 | {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, |
1e381f60 | 2039 | MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, |
26092bf5 | 2040 | {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | |
8dc0aa8c | 2041 | EXT4_MOUNT_JOURNAL_CHECKSUM), |
1e381f60 | 2042 | MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, |
8dc0aa8c | 2043 | {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, |
26092bf5 TT |
2044 | {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR}, |
2045 | {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR}, | |
2046 | {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR}, | |
8dc0aa8c | 2047 | {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, |
7915a861 | 2048 | MOPT_NO_EXT2}, |
8dc0aa8c | 2049 | {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, |
7915a861 | 2050 | MOPT_NO_EXT2}, |
26092bf5 TT |
2051 | {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, |
2052 | {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, | |
2053 | {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, | |
2054 | {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, | |
2055 | {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, | |
2056 | {Opt_commit, 0, MOPT_GTE0}, | |
2057 | {Opt_max_batch_time, 0, MOPT_GTE0}, | |
2058 | {Opt_min_batch_time, 0, MOPT_GTE0}, | |
2059 | {Opt_inode_readahead_blks, 0, MOPT_GTE0}, | |
2060 | {Opt_init_itable, 0, MOPT_GTE0}, | |
9cb20f94 IW |
2061 | {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP}, |
2062 | {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS, | |
2063 | MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP}, | |
2064 | {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE, | |
2065 | MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP}, | |
2066 | {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER, | |
2067 | MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP}, | |
26092bf5 | 2068 | {Opt_stripe, 0, MOPT_GTE0}, |
0efb3b23 JK |
2069 | {Opt_resuid, 0, MOPT_GTE0}, |
2070 | {Opt_resgid, 0, MOPT_GTE0}, | |
5ba92bcf CM |
2071 | {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0}, |
2072 | {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING}, | |
2073 | {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0}, | |
8dc0aa8c TT |
2074 | {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, |
2075 | {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, | |
2076 | {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, | |
2077 | MOPT_NO_EXT2 | MOPT_DATAJ}, | |
26092bf5 TT |
2078 | {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, |
2079 | {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, | |
03010a33 | 2080 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
26092bf5 TT |
2081 | {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, |
2082 | {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, | |
ac27a0ec | 2083 | #else |
26092bf5 TT |
2084 | {Opt_acl, 0, MOPT_NOSUPPORT}, |
2085 | {Opt_noacl, 0, MOPT_NOSUPPORT}, | |
ac27a0ec | 2086 | #endif |
26092bf5 TT |
2087 | {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, |
2088 | {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, | |
670e9875 | 2089 | {Opt_debug_want_extra_isize, 0, MOPT_GTE0}, |
26092bf5 TT |
2090 | {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, |
2091 | {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, | |
2092 | MOPT_SET | MOPT_Q}, | |
2093 | {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, | |
2094 | MOPT_SET | MOPT_Q}, | |
49da9392 JK |
2095 | {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA, |
2096 | MOPT_SET | MOPT_Q}, | |
26092bf5 | 2097 | {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | |
49da9392 JK |
2098 | EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), |
2099 | MOPT_CLEAR | MOPT_Q}, | |
174fe5ba KX |
2100 | {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING}, |
2101 | {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING}, | |
26092bf5 TT |
2102 | {Opt_offusrjquota, 0, MOPT_Q}, |
2103 | {Opt_offgrpjquota, 0, MOPT_Q}, | |
2104 | {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, | |
2105 | {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT}, | |
2106 | {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT}, | |
df981d03 | 2107 | {Opt_max_dir_size_kb, 0, MOPT_GTE0}, |
ed318a6c | 2108 | {Opt_test_dummy_encryption, 0, MOPT_STRING}, |
cdb7ee4c | 2109 | {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET}, |
21175ca4 | 2110 | {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS, |
3d392b26 | 2111 | MOPT_SET}, |
196e402a | 2112 | {Opt_mb_optimize_scan, EXT4_MOUNT2_MB_OPTIMIZE_SCAN, MOPT_GTE0}, |
99c880de | 2113 | #ifdef CONFIG_EXT4_DEBUG |
0f0672ff HS |
2114 | {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT, |
2115 | MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY}, | |
8016e29f HS |
2116 | {Opt_fc_debug_max_replay, 0, MOPT_GTE0}, |
2117 | #endif | |
26092bf5 TT |
2118 | {Opt_err, 0, 0} |
2119 | }; | |
2120 | ||
c83ad55e GKB |
2121 | #ifdef CONFIG_UNICODE |
2122 | static const struct ext4_sb_encodings { | |
2123 | __u16 magic; | |
2124 | char *name; | |
2125 | char *version; | |
2126 | } ext4_sb_encoding_map[] = { | |
2127 | {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"}, | |
2128 | }; | |
2129 | ||
2130 | static int ext4_sb_read_encoding(const struct ext4_super_block *es, | |
2131 | const struct ext4_sb_encodings **encoding, | |
2132 | __u16 *flags) | |
2133 | { | |
2134 | __u16 magic = le16_to_cpu(es->s_encoding); | |
2135 | int i; | |
2136 | ||
2137 | for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++) | |
2138 | if (magic == ext4_sb_encoding_map[i].magic) | |
2139 | break; | |
2140 | ||
2141 | if (i >= ARRAY_SIZE(ext4_sb_encoding_map)) | |
2142 | return -EINVAL; | |
2143 | ||
2144 | *encoding = &ext4_sb_encoding_map[i]; | |
2145 | *flags = le16_to_cpu(es->s_encoding_flags); | |
2146 | ||
2147 | return 0; | |
2148 | } | |
2149 | #endif | |
2150 | ||
6e47a3cc | 2151 | static int ext4_set_test_dummy_encryption(struct super_block *sb, char *arg) |
ed318a6c EB |
2152 | { |
2153 | #ifdef CONFIG_FS_ENCRYPTION | |
2154 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
2155 | int err; | |
2156 | ||
6e47a3cc | 2157 | err = fscrypt_set_test_dummy_encryption(sb, arg, |
ac4acb1f | 2158 | &sbi->s_dummy_enc_policy); |
ed318a6c | 2159 | if (err) { |
6e47a3cc LC |
2160 | ext4_msg(sb, KERN_WARNING, |
2161 | "Error while setting test dummy encryption [%d]", err); | |
2162 | return err; | |
ed318a6c EB |
2163 | } |
2164 | ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled"); | |
ed318a6c | 2165 | #endif |
6e47a3cc | 2166 | return 0; |
ed318a6c EB |
2167 | } |
2168 | ||
6e47a3cc LC |
2169 | #define EXT4_SPEC_JQUOTA (1 << 0) |
2170 | #define EXT4_SPEC_JQFMT (1 << 1) | |
2171 | #define EXT4_SPEC_DATAJ (1 << 2) | |
2172 | #define EXT4_SPEC_SB_BLOCK (1 << 3) | |
2173 | #define EXT4_SPEC_JOURNAL_DEV (1 << 4) | |
2174 | #define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5) | |
2175 | #define EXT4_SPEC_DUMMY_ENCRYPTION (1 << 6) | |
2176 | #define EXT4_SPEC_s_want_extra_isize (1 << 7) | |
2177 | #define EXT4_SPEC_s_max_batch_time (1 << 8) | |
2178 | #define EXT4_SPEC_s_min_batch_time (1 << 9) | |
2179 | #define EXT4_SPEC_s_inode_readahead_blks (1 << 10) | |
2180 | #define EXT4_SPEC_s_li_wait_mult (1 << 11) | |
2181 | #define EXT4_SPEC_s_max_dir_size_kb (1 << 12) | |
2182 | #define EXT4_SPEC_s_stripe (1 << 13) | |
2183 | #define EXT4_SPEC_s_resuid (1 << 14) | |
2184 | #define EXT4_SPEC_s_resgid (1 << 15) | |
2185 | #define EXT4_SPEC_s_commit_interval (1 << 16) | |
2186 | #define EXT4_SPEC_s_fc_debug_max_replay (1 << 17) | |
2187 | ||
461c3af0 | 2188 | struct ext4_fs_context { |
e6e268cb | 2189 | char *s_qf_names[EXT4_MAXQUOTAS]; |
6e47a3cc | 2190 | char *test_dummy_enc_arg; |
e6e268cb | 2191 | int s_jquota_fmt; /* Format of quota to use */ |
6e47a3cc LC |
2192 | int mb_optimize_scan; |
2193 | #ifdef CONFIG_EXT4_DEBUG | |
2194 | int s_fc_debug_max_replay; | |
2195 | #endif | |
e6e268cb | 2196 | unsigned short qname_spec; |
6e47a3cc LC |
2197 | unsigned long vals_s_flags; /* Bits to set in s_flags */ |
2198 | unsigned long mask_s_flags; /* Bits changed in s_flags */ | |
e6e268cb | 2199 | unsigned long journal_devnum; |
6e47a3cc LC |
2200 | unsigned long s_commit_interval; |
2201 | unsigned long s_stripe; | |
2202 | unsigned int s_inode_readahead_blks; | |
2203 | unsigned int s_want_extra_isize; | |
2204 | unsigned int s_li_wait_mult; | |
2205 | unsigned int s_max_dir_size_kb; | |
e6e268cb | 2206 | unsigned int journal_ioprio; |
6e47a3cc LC |
2207 | unsigned int vals_s_mount_opt; |
2208 | unsigned int mask_s_mount_opt; | |
2209 | unsigned int vals_s_mount_opt2; | |
2210 | unsigned int mask_s_mount_opt2; | |
2211 | unsigned int vals_s_mount_flags; | |
2212 | unsigned int mask_s_mount_flags; | |
b6bd2435 | 2213 | unsigned int opt_flags; /* MOPT flags */ |
6e47a3cc LC |
2214 | unsigned int spec; |
2215 | u32 s_max_batch_time; | |
2216 | u32 s_min_batch_time; | |
2217 | kuid_t s_resuid; | |
2218 | kgid_t s_resgid; | |
b237e304 HS |
2219 | }; |
2220 | ||
e6e268cb LC |
2221 | #ifdef CONFIG_QUOTA |
2222 | /* | |
2223 | * Note the name of the specified quota file. | |
2224 | */ | |
2225 | static int note_qf_name(struct fs_context *fc, int qtype, | |
2226 | struct fs_parameter *param) | |
2227 | { | |
2228 | struct ext4_fs_context *ctx = fc->fs_private; | |
2229 | char *qname; | |
2230 | ||
2231 | if (param->size < 1) { | |
2232 | ext4_msg(NULL, KERN_ERR, "Missing quota name"); | |
2233 | return -EINVAL; | |
2234 | } | |
2235 | if (strchr(param->string, '/')) { | |
2236 | ext4_msg(NULL, KERN_ERR, | |
2237 | "quotafile must be on filesystem root"); | |
2238 | return -EINVAL; | |
2239 | } | |
2240 | if (ctx->s_qf_names[qtype]) { | |
2241 | if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) { | |
2242 | ext4_msg(NULL, KERN_ERR, | |
2243 | "%s quota file already specified", | |
2244 | QTYPE2NAME(qtype)); | |
2245 | return -EINVAL; | |
2246 | } | |
2247 | return 0; | |
2248 | } | |
2249 | ||
2250 | qname = kmemdup_nul(param->string, param->size, GFP_KERNEL); | |
2251 | if (!qname) { | |
2252 | ext4_msg(NULL, KERN_ERR, | |
2253 | "Not enough memory for storing quotafile name"); | |
2254 | return -ENOMEM; | |
2255 | } | |
2256 | ctx->s_qf_names[qtype] = qname; | |
2257 | ctx->qname_spec |= 1 << qtype; | |
6e47a3cc | 2258 | ctx->spec |= EXT4_SPEC_JQUOTA; |
e6e268cb LC |
2259 | return 0; |
2260 | } | |
2261 | ||
2262 | /* | |
2263 | * Clear the name of the specified quota file. | |
2264 | */ | |
2265 | static int unnote_qf_name(struct fs_context *fc, int qtype) | |
2266 | { | |
2267 | struct ext4_fs_context *ctx = fc->fs_private; | |
2268 | ||
2269 | if (ctx->s_qf_names[qtype]) | |
2270 | kfree(ctx->s_qf_names[qtype]); | |
2271 | ||
2272 | ctx->s_qf_names[qtype] = NULL; | |
2273 | ctx->qname_spec |= 1 << qtype; | |
6e47a3cc | 2274 | ctx->spec |= EXT4_SPEC_JQUOTA; |
e6e268cb LC |
2275 | return 0; |
2276 | } | |
2277 | #endif | |
2278 | ||
6e47a3cc LC |
2279 | #define EXT4_SET_CTX(name) \ |
2280 | static inline void ctx_set_##name(struct ext4_fs_context *ctx, int flag)\ | |
2281 | { \ | |
2282 | ctx->mask_s_##name |= flag; \ | |
2283 | ctx->vals_s_##name |= flag; \ | |
2284 | } \ | |
2285 | static inline void ctx_clear_##name(struct ext4_fs_context *ctx, int flag)\ | |
2286 | { \ | |
2287 | ctx->mask_s_##name |= flag; \ | |
2288 | ctx->vals_s_##name &= ~flag; \ | |
2289 | } \ | |
2290 | static inline bool ctx_test_##name(struct ext4_fs_context *ctx, int flag)\ | |
2291 | { \ | |
2292 | return ((ctx->vals_s_##name & flag) != 0); \ | |
2293 | } \ | |
2294 | ||
2295 | EXT4_SET_CTX(flags); | |
2296 | EXT4_SET_CTX(mount_opt); | |
2297 | EXT4_SET_CTX(mount_opt2); | |
2298 | EXT4_SET_CTX(mount_flags); | |
2299 | ||
461c3af0 | 2300 | static int handle_mount_opt(struct fs_context *fc, struct fs_parameter *param) |
26092bf5 | 2301 | { |
461c3af0 | 2302 | struct ext4_fs_context *ctx = fc->fs_private; |
461c3af0 | 2303 | struct fs_parse_result result; |
26092bf5 | 2304 | const struct mount_opts *m; |
461c3af0 | 2305 | int is_remount; |
08cefc7a EB |
2306 | kuid_t uid; |
2307 | kgid_t gid; | |
461c3af0 LC |
2308 | int token; |
2309 | ||
2310 | token = fs_parse(fc, ext4_param_specs, param, &result); | |
2311 | if (token < 0) | |
2312 | return token; | |
2313 | is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE; | |
26092bf5 | 2314 | |
57f73c2c | 2315 | #ifdef CONFIG_QUOTA |
461c3af0 LC |
2316 | if (token == Opt_usrjquota) { |
2317 | if (!*param->string) | |
e6e268cb | 2318 | return unnote_qf_name(fc, USRQUOTA); |
461c3af0 | 2319 | else |
e6e268cb | 2320 | return note_qf_name(fc, USRQUOTA, param); |
461c3af0 LC |
2321 | } else if (token == Opt_grpjquota) { |
2322 | if (!*param->string) | |
e6e268cb | 2323 | return unnote_qf_name(fc, GRPQUOTA); |
461c3af0 | 2324 | else |
e6e268cb | 2325 | return note_qf_name(fc, GRPQUOTA, param); |
461c3af0 | 2326 | } |
57f73c2c | 2327 | #endif |
26092bf5 | 2328 | switch (token) { |
f7048605 TT |
2329 | case Opt_noacl: |
2330 | case Opt_nouser_xattr: | |
da812f61 | 2331 | ext4_msg(NULL, KERN_WARNING, deprecated_msg, param->key, "3.5"); |
f7048605 | 2332 | break; |
26092bf5 TT |
2333 | case Opt_sb: |
2334 | return 1; /* handled by get_sb_block() */ | |
2335 | case Opt_removed: | |
da812f61 | 2336 | ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option", |
461c3af0 | 2337 | param->key); |
26092bf5 | 2338 | return 1; |
26092bf5 | 2339 | case Opt_abort: |
6e47a3cc | 2340 | ctx_set_mount_flags(ctx, EXT4_MF_FS_ABORTED); |
26092bf5 TT |
2341 | return 1; |
2342 | case Opt_i_version: | |
6e47a3cc | 2343 | ctx_set_flags(ctx, SB_I_VERSION); |
26092bf5 | 2344 | return 1; |
a26f4992 | 2345 | case Opt_lazytime: |
6e47a3cc | 2346 | ctx_set_flags(ctx, SB_LAZYTIME); |
a26f4992 TT |
2347 | return 1; |
2348 | case Opt_nolazytime: | |
6e47a3cc | 2349 | ctx_clear_flags(ctx, SB_LAZYTIME); |
a26f4992 | 2350 | return 1; |
4f74d15f EB |
2351 | case Opt_inlinecrypt: |
2352 | #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT | |
6e47a3cc | 2353 | ctx_set_flags(ctx, SB_INLINECRYPT); |
4f74d15f | 2354 | #else |
da812f61 | 2355 | ext4_msg(NULL, KERN_ERR, "inline encryption not supported"); |
4f74d15f EB |
2356 | #endif |
2357 | return 1; | |
461c3af0 LC |
2358 | case Opt_errors: |
2359 | case Opt_data: | |
2360 | case Opt_data_err: | |
2361 | case Opt_jqfmt: | |
2362 | case Opt_dax_type: | |
2363 | token = result.uint_32; | |
26092bf5 TT |
2364 | } |
2365 | ||
5f3633e3 JK |
2366 | for (m = ext4_mount_opts; m->token != Opt_err; m++) |
2367 | if (token == m->token) | |
2368 | break; | |
2369 | ||
b6bd2435 LC |
2370 | ctx->opt_flags |= m->flags; |
2371 | ||
5f3633e3 | 2372 | if (m->token == Opt_err) { |
da812f61 | 2373 | ext4_msg(NULL, KERN_ERR, "Unrecognized mount option \"%s\" " |
461c3af0 | 2374 | "or missing value", param->key); |
da812f61 | 2375 | return -EINVAL; |
5f3633e3 JK |
2376 | } |
2377 | ||
c93cf2d7 DM |
2378 | if (m->flags & MOPT_EXPLICIT) { |
2379 | if (m->mount_opt & EXT4_MOUNT_DELALLOC) { | |
6e47a3cc | 2380 | ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC); |
1e381f60 | 2381 | } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { |
6e47a3cc LC |
2382 | ctx_set_mount_opt2(ctx, |
2383 | EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM); | |
c93cf2d7 | 2384 | } else |
da812f61 | 2385 | return -EINVAL; |
c93cf2d7 | 2386 | } |
5f3633e3 | 2387 | if (m->flags & MOPT_CLEAR_ERR) |
6e47a3cc | 2388 | ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK); |
5f3633e3 JK |
2389 | |
2390 | if (m->flags & MOPT_NOSUPPORT) { | |
da812f61 | 2391 | ext4_msg(NULL, KERN_ERR, "%s option not supported", |
461c3af0 | 2392 | param->key); |
5f3633e3 | 2393 | } else if (token == Opt_commit) { |
461c3af0 | 2394 | if (result.uint_32 == 0) |
6e47a3cc | 2395 | ctx->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE; |
461c3af0 | 2396 | else if (result.uint_32 > INT_MAX / HZ) { |
da812f61 | 2397 | ext4_msg(NULL, KERN_ERR, |
9ba55543 | 2398 | "Invalid commit interval %d, " |
2399 | "must be smaller than %d", | |
461c3af0 | 2400 | result.uint_32, INT_MAX / HZ); |
da812f61 | 2401 | return -EINVAL; |
9ba55543 | 2402 | } |
6e47a3cc LC |
2403 | ctx->s_commit_interval = HZ * result.uint_32; |
2404 | ctx->spec |= EXT4_SPEC_s_commit_interval; | |
670e9875 | 2405 | } else if (token == Opt_debug_want_extra_isize) { |
6e47a3cc | 2406 | if ((result.uint_32 & 1) || (result.uint_32 < 4)) { |
da812f61 | 2407 | ext4_msg(NULL, KERN_ERR, |
461c3af0 | 2408 | "Invalid want_extra_isize %d", result.uint_32); |
da812f61 | 2409 | return -EINVAL; |
9803387c | 2410 | } |
6e47a3cc LC |
2411 | ctx->s_want_extra_isize = result.uint_32; |
2412 | ctx->spec |= EXT4_SPEC_s_want_extra_isize; | |
5f3633e3 | 2413 | } else if (token == Opt_max_batch_time) { |
6e47a3cc LC |
2414 | ctx->s_max_batch_time = result.uint_32; |
2415 | ctx->spec |= EXT4_SPEC_s_max_batch_time; | |
5f3633e3 | 2416 | } else if (token == Opt_min_batch_time) { |
6e47a3cc LC |
2417 | ctx->s_min_batch_time = result.uint_32; |
2418 | ctx->spec |= EXT4_SPEC_s_min_batch_time; | |
5f3633e3 | 2419 | } else if (token == Opt_inode_readahead_blks) { |
461c3af0 LC |
2420 | if (result.uint_32 && |
2421 | (result.uint_32 > (1 << 30) || | |
2422 | !is_power_of_2(result.uint_32))) { | |
da812f61 | 2423 | ext4_msg(NULL, KERN_ERR, |
e33e60ea JK |
2424 | "EXT4-fs: inode_readahead_blks must be " |
2425 | "0 or a power of 2 smaller than 2^31"); | |
da812f61 | 2426 | return -EINVAL; |
5f3633e3 | 2427 | } |
6e47a3cc LC |
2428 | ctx->s_inode_readahead_blks = result.uint_32; |
2429 | ctx->spec |= EXT4_SPEC_s_inode_readahead_blks; | |
5f3633e3 | 2430 | } else if (token == Opt_init_itable) { |
6e47a3cc LC |
2431 | ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE); |
2432 | ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; | |
461c3af0 | 2433 | if (param->type == fs_value_is_string) |
6e47a3cc LC |
2434 | ctx->s_li_wait_mult = result.uint_32; |
2435 | ctx->spec |= EXT4_SPEC_s_li_wait_mult; | |
5f3633e3 | 2436 | } else if (token == Opt_max_dir_size_kb) { |
6e47a3cc LC |
2437 | ctx->s_max_dir_size_kb = result.uint_32; |
2438 | ctx->spec |= EXT4_SPEC_s_max_dir_size_kb; | |
8016e29f HS |
2439 | #ifdef CONFIG_EXT4_DEBUG |
2440 | } else if (token == Opt_fc_debug_max_replay) { | |
6e47a3cc LC |
2441 | ctx->s_fc_debug_max_replay = result.uint_32; |
2442 | ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay; | |
8016e29f | 2443 | #endif |
5f3633e3 | 2444 | } else if (token == Opt_stripe) { |
6e47a3cc LC |
2445 | ctx->s_stripe = result.uint_32; |
2446 | ctx->spec |= EXT4_SPEC_s_stripe; | |
5f3633e3 | 2447 | } else if (token == Opt_resuid) { |
461c3af0 | 2448 | uid = make_kuid(current_user_ns(), result.uint_32); |
5f3633e3 | 2449 | if (!uid_valid(uid)) { |
da812f61 | 2450 | ext4_msg(NULL, KERN_ERR, "Invalid uid value %d", |
461c3af0 | 2451 | result.uint_32); |
da812f61 | 2452 | return -EINVAL; |
26092bf5 | 2453 | } |
6e47a3cc LC |
2454 | ctx->s_resuid = uid; |
2455 | ctx->spec |= EXT4_SPEC_s_resuid; | |
5f3633e3 | 2456 | } else if (token == Opt_resgid) { |
461c3af0 | 2457 | gid = make_kgid(current_user_ns(), result.uint_32); |
5f3633e3 | 2458 | if (!gid_valid(gid)) { |
da812f61 | 2459 | ext4_msg(NULL, KERN_ERR, "Invalid gid value %d", |
461c3af0 | 2460 | result.uint_32); |
da812f61 | 2461 | return -EINVAL; |
5f3633e3 | 2462 | } |
6e47a3cc LC |
2463 | ctx->s_resgid = gid; |
2464 | ctx->spec |= EXT4_SPEC_s_resgid; | |
5f3633e3 JK |
2465 | } else if (token == Opt_journal_dev) { |
2466 | if (is_remount) { | |
da812f61 | 2467 | ext4_msg(NULL, KERN_ERR, |
5f3633e3 | 2468 | "Cannot specify journal on remount"); |
da812f61 | 2469 | return -EINVAL; |
5f3633e3 | 2470 | } |
461c3af0 | 2471 | ctx->journal_devnum = result.uint_32; |
6e47a3cc | 2472 | ctx->spec |= EXT4_SPEC_JOURNAL_DEV; |
ad4eec61 | 2473 | } else if (token == Opt_journal_path) { |
ad4eec61 ES |
2474 | struct inode *journal_inode; |
2475 | struct path path; | |
2476 | int error; | |
2477 | ||
2478 | if (is_remount) { | |
da812f61 | 2479 | ext4_msg(NULL, KERN_ERR, |
ad4eec61 | 2480 | "Cannot specify journal on remount"); |
da812f61 | 2481 | return -EINVAL; |
ad4eec61 | 2482 | } |
ad4eec61 | 2483 | |
461c3af0 | 2484 | error = fs_lookup_param(fc, param, 1, &path); |
ad4eec61 | 2485 | if (error) { |
da812f61 | 2486 | ext4_msg(NULL, KERN_ERR, "error: could not find " |
461c3af0 | 2487 | "journal device path"); |
da812f61 | 2488 | return -EINVAL; |
ad4eec61 ES |
2489 | } |
2490 | ||
2b0143b5 | 2491 | journal_inode = d_inode(path.dentry); |
461c3af0 | 2492 | ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev); |
6e47a3cc | 2493 | ctx->spec |= EXT4_SPEC_JOURNAL_DEV; |
ad4eec61 | 2494 | path_put(&path); |
5f3633e3 | 2495 | } else if (token == Opt_journal_ioprio) { |
461c3af0 | 2496 | if (result.uint_32 > 7) { |
da812f61 | 2497 | ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority" |
5f3633e3 | 2498 | " (must be 0-7)"); |
da812f61 | 2499 | return -EINVAL; |
5f3633e3 | 2500 | } |
461c3af0 LC |
2501 | ctx->journal_ioprio = |
2502 | IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32); | |
6e47a3cc | 2503 | ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO; |
6ddb2447 | 2504 | } else if (token == Opt_test_dummy_encryption) { |
6e47a3cc LC |
2505 | #ifdef CONFIG_FS_ENCRYPTION |
2506 | if (param->type == fs_value_is_flag) { | |
2507 | ctx->spec |= EXT4_SPEC_DUMMY_ENCRYPTION; | |
2508 | ctx->test_dummy_enc_arg = NULL; | |
2509 | return 1; | |
ac27a0ec | 2510 | } |
6e47a3cc LC |
2511 | if (*param->string && |
2512 | !(!strcmp(param->string, "v1") || | |
2513 | !strcmp(param->string, "v2"))) { | |
2514 | ext4_msg(NULL, KERN_WARNING, | |
2515 | "Value of option \"%s\" is unrecognized", | |
2516 | param->key); | |
2517 | return -EINVAL; | |
2518 | } | |
2519 | ctx->spec |= EXT4_SPEC_DUMMY_ENCRYPTION; | |
2520 | ctx->test_dummy_enc_arg = kmemdup_nul(param->string, param->size, | |
2521 | GFP_KERNEL); | |
2522 | #else | |
2523 | ext4_msg(NULL, KERN_WARNING, | |
2524 | "Test dummy encryption mount option ignored"); | |
2525 | #endif | |
2526 | } else if (m->flags & MOPT_DATAJ) { | |
2527 | ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS); | |
2528 | ctx_set_mount_opt(ctx, m->mount_opt); | |
2529 | ctx->spec |= EXT4_SPEC_DATAJ; | |
5f3633e3 JK |
2530 | #ifdef CONFIG_QUOTA |
2531 | } else if (m->flags & MOPT_QFMT) { | |
e6e268cb | 2532 | ctx->s_jquota_fmt = m->mount_opt; |
6e47a3cc | 2533 | ctx->spec |= EXT4_SPEC_JQFMT; |
923ae0ff | 2534 | #endif |
9cb20f94 IW |
2535 | } else if (token == Opt_dax || token == Opt_dax_always || |
2536 | token == Opt_dax_inode || token == Opt_dax_never) { | |
ef83b6e8 | 2537 | #ifdef CONFIG_FS_DAX |
9cb20f94 IW |
2538 | switch (token) { |
2539 | case Opt_dax: | |
2540 | case Opt_dax_always: | |
6e47a3cc LC |
2541 | ctx_set_mount_opt(ctx, m->mount_opt); |
2542 | ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER); | |
9cb20f94 IW |
2543 | break; |
2544 | case Opt_dax_never: | |
6e47a3cc LC |
2545 | ctx_set_mount_opt2(ctx, m->mount_opt); |
2546 | ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS); | |
9cb20f94 IW |
2547 | break; |
2548 | case Opt_dax_inode: | |
6e47a3cc LC |
2549 | ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS); |
2550 | ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER); | |
9cb20f94 | 2551 | /* Strictly for printing options */ |
6e47a3cc | 2552 | ctx_set_mount_opt2(ctx, m->mount_opt); |
9cb20f94 IW |
2553 | break; |
2554 | } | |
ef83b6e8 | 2555 | #else |
da812f61 | 2556 | ext4_msg(NULL, KERN_INFO, "dax option not supported"); |
6e47a3cc LC |
2557 | ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER); |
2558 | ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS); | |
da812f61 | 2559 | return -EINVAL; |
5f3633e3 | 2560 | #endif |
7915a861 | 2561 | } else if (token == Opt_data_err_abort) { |
6e47a3cc | 2562 | ctx_set_mount_opt(ctx, m->mount_opt); |
7915a861 | 2563 | } else if (token == Opt_data_err_ignore) { |
6e47a3cc | 2564 | ctx_clear_mount_opt(ctx, m->mount_opt); |
196e402a | 2565 | } else if (token == Opt_mb_optimize_scan) { |
461c3af0 | 2566 | if (result.int_32 != 0 && result.int_32 != 1) { |
da812f61 | 2567 | ext4_msg(NULL, KERN_WARNING, |
196e402a | 2568 | "mb_optimize_scan should be set to 0 or 1."); |
da812f61 | 2569 | return -EINVAL; |
196e402a | 2570 | } |
461c3af0 | 2571 | ctx->mb_optimize_scan = result.int_32; |
5f3633e3 | 2572 | } else { |
461c3af0 LC |
2573 | unsigned int set = 0; |
2574 | ||
2575 | if ((param->type == fs_value_is_flag) || | |
2576 | result.uint_32 > 0) | |
2577 | set = 1; | |
2578 | ||
5f3633e3 | 2579 | if (m->flags & MOPT_CLEAR) |
461c3af0 | 2580 | set = !set; |
5f3633e3 | 2581 | else if (unlikely(!(m->flags & MOPT_SET))) { |
da812f61 | 2582 | ext4_msg(NULL, KERN_WARNING, |
461c3af0 LC |
2583 | "buggy handling of option %s", |
2584 | param->key); | |
5f3633e3 | 2585 | WARN_ON(1); |
da812f61 | 2586 | return -EINVAL; |
5f3633e3 | 2587 | } |
995a3ed6 | 2588 | if (m->flags & MOPT_2) { |
461c3af0 | 2589 | if (set != 0) |
6e47a3cc | 2590 | ctx_set_mount_opt2(ctx, m->mount_opt); |
995a3ed6 | 2591 | else |
6e47a3cc | 2592 | ctx_clear_mount_opt2(ctx, m->mount_opt); |
995a3ed6 | 2593 | } else { |
461c3af0 | 2594 | if (set != 0) |
6e47a3cc | 2595 | ctx_set_mount_opt(ctx, m->mount_opt); |
995a3ed6 | 2596 | else |
6e47a3cc | 2597 | ctx_clear_mount_opt(ctx, m->mount_opt); |
995a3ed6 | 2598 | } |
26092bf5 | 2599 | } |
5f3633e3 | 2600 | return 1; |
26092bf5 TT |
2601 | } |
2602 | ||
2603 | static int parse_options(char *options, struct super_block *sb, | |
e6e268cb | 2604 | struct ext4_fs_context *ctx, |
26092bf5 TT |
2605 | int is_remount) |
2606 | { | |
461c3af0 LC |
2607 | struct fs_parameter param; |
2608 | struct fs_context fc; | |
2609 | int ret; | |
2610 | char *key; | |
26092bf5 TT |
2611 | |
2612 | if (!options) | |
2613 | return 1; | |
2614 | ||
461c3af0 | 2615 | memset(&fc, 0, sizeof(fc)); |
e6e268cb | 2616 | fc.fs_private = ctx; |
461c3af0 LC |
2617 | fc.s_fs_info = EXT4_SB(sb); |
2618 | ||
2619 | if (is_remount) | |
2620 | fc.purpose = FS_CONTEXT_FOR_RECONFIGURE; | |
2621 | ||
2622 | while ((key = strsep(&options, ",")) != NULL) { | |
2623 | if (*key) { | |
2624 | size_t v_len = 0; | |
2625 | char *value = strchr(key, '='); | |
2626 | ||
2627 | param.type = fs_value_is_flag; | |
2628 | param.string = NULL; | |
2629 | ||
2630 | if (value) { | |
2631 | if (value == key) | |
2632 | continue; | |
2633 | ||
2634 | *value++ = 0; | |
2635 | v_len = strlen(value); | |
2636 | param.string = kmemdup_nul(value, v_len, | |
2637 | GFP_KERNEL); | |
2638 | if (!param.string) | |
2639 | return 0; | |
2640 | param.type = fs_value_is_string; | |
2641 | } | |
2642 | ||
2643 | param.key = key; | |
2644 | param.size = v_len; | |
2645 | ||
2646 | ret = handle_mount_opt(&fc, ¶m); | |
2647 | if (param.string) | |
2648 | kfree(param.string); | |
2649 | if (ret < 0) | |
2650 | return 0; | |
2651 | } | |
ac27a0ec | 2652 | } |
461c3af0 | 2653 | |
da812f61 LC |
2654 | ret = ext4_validate_options(&fc); |
2655 | if (ret < 0) | |
2656 | return 0; | |
2657 | ||
b6bd2435 | 2658 | ret = ext4_check_opt_consistency(&fc, sb); |
e6e268cb LC |
2659 | if (ret < 0) |
2660 | return 0; | |
2661 | ||
6e47a3cc LC |
2662 | ret = ext4_apply_options(&fc, sb); |
2663 | if (ret < 0) | |
2664 | return 0; | |
e6e268cb | 2665 | |
da812f61 | 2666 | return 1; |
4c94bff9 LC |
2667 | } |
2668 | ||
e6e268cb LC |
2669 | static void ext4_apply_quota_options(struct fs_context *fc, |
2670 | struct super_block *sb) | |
2671 | { | |
2672 | #ifdef CONFIG_QUOTA | |
6e47a3cc | 2673 | bool quota_feature = ext4_has_feature_quota(sb); |
e6e268cb LC |
2674 | struct ext4_fs_context *ctx = fc->fs_private; |
2675 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
2676 | char *qname; | |
2677 | int i; | |
2678 | ||
6e47a3cc LC |
2679 | if (quota_feature) |
2680 | return; | |
2681 | ||
2682 | if (ctx->spec & EXT4_SPEC_JQUOTA) { | |
2683 | for (i = 0; i < EXT4_MAXQUOTAS; i++) { | |
2684 | if (!(ctx->qname_spec & (1 << i))) | |
2685 | continue; | |
2686 | ||
2687 | qname = ctx->s_qf_names[i]; /* May be NULL */ | |
2688 | ctx->s_qf_names[i] = NULL; | |
2689 | kfree(sbi->s_qf_names[i]); | |
2690 | rcu_assign_pointer(sbi->s_qf_names[i], qname); | |
2691 | set_opt(sb, QUOTA); | |
2692 | } | |
e6e268cb | 2693 | } |
6e47a3cc LC |
2694 | |
2695 | if (ctx->spec & EXT4_SPEC_JQFMT) | |
2696 | sbi->s_jquota_fmt = ctx->s_jquota_fmt; | |
e6e268cb LC |
2697 | #endif |
2698 | } | |
2699 | ||
2700 | /* | |
2701 | * Check quota settings consistency. | |
2702 | */ | |
2703 | static int ext4_check_quota_consistency(struct fs_context *fc, | |
2704 | struct super_block *sb) | |
2705 | { | |
2706 | #ifdef CONFIG_QUOTA | |
2707 | struct ext4_fs_context *ctx = fc->fs_private; | |
2708 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
2709 | bool quota_feature = ext4_has_feature_quota(sb); | |
2710 | bool quota_loaded = sb_any_quota_loaded(sb); | |
6e47a3cc LC |
2711 | bool usr_qf_name, grp_qf_name, usrquota, grpquota; |
2712 | int quota_flags, i; | |
2713 | ||
2714 | /* | |
2715 | * We do the test below only for project quotas. 'usrquota' and | |
2716 | * 'grpquota' mount options are allowed even without quota feature | |
2717 | * to support legacy quotas in quota files. | |
2718 | */ | |
2719 | if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) && | |
2720 | !ext4_has_feature_project(sb)) { | |
2721 | ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. " | |
2722 | "Cannot enable project quota enforcement."); | |
2723 | return -EINVAL; | |
2724 | } | |
e6e268cb | 2725 | |
6e47a3cc LC |
2726 | quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | |
2727 | EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA; | |
2728 | if (quota_loaded && | |
2729 | ctx->mask_s_mount_opt & quota_flags && | |
2730 | !ctx_test_mount_opt(ctx, quota_flags)) | |
2731 | goto err_quota_change; | |
2732 | ||
2733 | if (ctx->spec & EXT4_SPEC_JQUOTA) { | |
e6e268cb LC |
2734 | |
2735 | for (i = 0; i < EXT4_MAXQUOTAS; i++) { | |
2736 | if (!(ctx->qname_spec & (1 << i))) | |
2737 | continue; | |
2738 | ||
6e47a3cc LC |
2739 | if (quota_loaded && |
2740 | !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i]) | |
e6e268cb LC |
2741 | goto err_jquota_change; |
2742 | ||
2743 | if (sbi->s_qf_names[i] && ctx->s_qf_names[i] && | |
2744 | strcmp(sbi->s_qf_names[i], | |
2745 | ctx->s_qf_names[i]) != 0) | |
2746 | goto err_jquota_specified; | |
2747 | } | |
6e47a3cc LC |
2748 | |
2749 | if (quota_feature) { | |
2750 | ext4_msg(NULL, KERN_INFO, | |
2751 | "Journaled quota options ignored when " | |
2752 | "QUOTA feature is enabled"); | |
2753 | return 0; | |
2754 | } | |
e6e268cb LC |
2755 | } |
2756 | ||
6e47a3cc | 2757 | if (ctx->spec & EXT4_SPEC_JQFMT) { |
e6e268cb | 2758 | if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded) |
6e47a3cc | 2759 | goto err_jquota_change; |
e6e268cb LC |
2760 | if (quota_feature) { |
2761 | ext4_msg(NULL, KERN_INFO, "Quota format mount options " | |
2762 | "ignored when QUOTA feature is enabled"); | |
2763 | return 0; | |
2764 | } | |
2765 | } | |
6e47a3cc LC |
2766 | |
2767 | /* Make sure we don't mix old and new quota format */ | |
2768 | usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) || | |
2769 | ctx->s_qf_names[USRQUOTA]); | |
2770 | grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) || | |
2771 | ctx->s_qf_names[GRPQUOTA]); | |
2772 | ||
2773 | usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) || | |
2774 | test_opt(sb, USRQUOTA)); | |
2775 | ||
2776 | grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) || | |
2777 | test_opt(sb, GRPQUOTA)); | |
2778 | ||
2779 | if (usr_qf_name) { | |
2780 | ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA); | |
2781 | usrquota = false; | |
2782 | } | |
2783 | if (grp_qf_name) { | |
2784 | ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA); | |
2785 | grpquota = false; | |
2786 | } | |
2787 | ||
2788 | if (usr_qf_name || grp_qf_name) { | |
2789 | if (usrquota || grpquota) { | |
2790 | ext4_msg(NULL, KERN_ERR, "old and new quota " | |
2791 | "format mixing"); | |
2792 | return -EINVAL; | |
2793 | } | |
2794 | ||
2795 | if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) { | |
2796 | ext4_msg(NULL, KERN_ERR, "journaled quota format " | |
2797 | "not specified"); | |
2798 | return -EINVAL; | |
2799 | } | |
2800 | } | |
2801 | ||
e6e268cb LC |
2802 | return 0; |
2803 | ||
2804 | err_quota_change: | |
2805 | ext4_msg(NULL, KERN_ERR, | |
2806 | "Cannot change quota options when quota turned on"); | |
2807 | return -EINVAL; | |
2808 | err_jquota_change: | |
2809 | ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota " | |
2810 | "options when quota turned on"); | |
2811 | return -EINVAL; | |
2812 | err_jquota_specified: | |
2813 | ext4_msg(NULL, KERN_ERR, "%s quota file already specified", | |
2814 | QTYPE2NAME(i)); | |
2815 | return -EINVAL; | |
e6e268cb LC |
2816 | #else |
2817 | return 0; | |
2818 | #endif | |
2819 | } | |
2820 | ||
b6bd2435 LC |
2821 | static int ext4_check_opt_consistency(struct fs_context *fc, |
2822 | struct super_block *sb) | |
2823 | { | |
2824 | struct ext4_fs_context *ctx = fc->fs_private; | |
6e47a3cc LC |
2825 | struct ext4_sb_info *sbi = fc->s_fs_info; |
2826 | int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE; | |
b6bd2435 LC |
2827 | |
2828 | if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { | |
2829 | ext4_msg(NULL, KERN_ERR, | |
2830 | "Mount option(s) incompatible with ext2"); | |
2831 | return -EINVAL; | |
2832 | } | |
2833 | if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { | |
2834 | ext4_msg(NULL, KERN_ERR, | |
2835 | "Mount option(s) incompatible with ext3"); | |
2836 | return -EINVAL; | |
2837 | } | |
2838 | ||
6e47a3cc LC |
2839 | if (ctx->s_want_extra_isize > |
2840 | (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) { | |
2841 | ext4_msg(NULL, KERN_ERR, | |
2842 | "Invalid want_extra_isize %d", | |
2843 | ctx->s_want_extra_isize); | |
2844 | return -EINVAL; | |
2845 | } | |
2846 | ||
2847 | if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DIOREAD_NOLOCK)) { | |
2848 | int blocksize = | |
2849 | BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); | |
2850 | if (blocksize < PAGE_SIZE) | |
2851 | ext4_msg(NULL, KERN_WARNING, "Warning: mounting with an " | |
2852 | "experimental mount option 'dioread_nolock' " | |
2853 | "for blocksize < PAGE_SIZE"); | |
2854 | } | |
2855 | ||
2856 | #ifdef CONFIG_FS_ENCRYPTION | |
2857 | /* | |
2858 | * This mount option is just for testing, and it's not worthwhile to | |
2859 | * implement the extra complexity (e.g. RCU protection) that would be | |
2860 | * needed to allow it to be set or changed during remount. We do allow | |
2861 | * it to be specified during remount, but only if there is no change. | |
2862 | */ | |
2863 | if ((ctx->spec & EXT4_SPEC_DUMMY_ENCRYPTION) && | |
2864 | is_remount && !sbi->s_dummy_enc_policy.policy) { | |
2865 | ext4_msg(NULL, KERN_WARNING, | |
2866 | "Can't set test_dummy_encryption on remount"); | |
2867 | return -1; | |
2868 | } | |
2869 | #endif | |
2870 | ||
2871 | if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) { | |
2872 | if (!sbi->s_journal) { | |
2873 | ext4_msg(NULL, KERN_WARNING, | |
2874 | "Remounting file system with no journal " | |
2875 | "so ignoring journalled data option"); | |
2876 | ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS); | |
2877 | } else if (ctx->mask_s_mount_opt & EXT4_MOUNT_DATA_FLAGS) { | |
2878 | ext4_msg(NULL, KERN_ERR, "Cannot change data mode " | |
2879 | "on remount"); | |
2880 | return -EINVAL; | |
2881 | } | |
2882 | } | |
2883 | ||
2884 | if (is_remount) { | |
2885 | if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) && | |
2886 | (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) { | |
2887 | ext4_msg(NULL, KERN_ERR, "can't mount with " | |
2888 | "both data=journal and dax"); | |
2889 | return -EINVAL; | |
2890 | } | |
2891 | ||
2892 | if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) && | |
2893 | (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || | |
2894 | (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) { | |
2895 | fail_dax_change_remount: | |
2896 | ext4_msg(NULL, KERN_ERR, "can't change " | |
2897 | "dax mount option while remounting"); | |
2898 | return -EINVAL; | |
2899 | } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) && | |
2900 | (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || | |
2901 | (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) { | |
2902 | goto fail_dax_change_remount; | |
2903 | } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) && | |
2904 | ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || | |
2905 | (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || | |
2906 | !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) { | |
2907 | goto fail_dax_change_remount; | |
2908 | } | |
2909 | } | |
2910 | ||
b6bd2435 LC |
2911 | return ext4_check_quota_consistency(fc, sb); |
2912 | } | |
2913 | ||
6e47a3cc | 2914 | static int ext4_apply_options(struct fs_context *fc, struct super_block *sb) |
4c94bff9 | 2915 | { |
6e47a3cc | 2916 | struct ext4_fs_context *ctx = fc->fs_private; |
da812f61 | 2917 | struct ext4_sb_info *sbi = fc->s_fs_info; |
6e47a3cc LC |
2918 | int ret = 0; |
2919 | ||
2920 | sbi->s_mount_opt &= ~ctx->mask_s_mount_opt; | |
2921 | sbi->s_mount_opt |= ctx->vals_s_mount_opt; | |
2922 | sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2; | |
2923 | sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2; | |
2924 | sbi->s_mount_flags &= ~ctx->mask_s_mount_flags; | |
2925 | sbi->s_mount_flags |= ctx->vals_s_mount_flags; | |
2926 | sb->s_flags &= ~ctx->mask_s_flags; | |
2927 | sb->s_flags |= ctx->vals_s_flags; | |
2928 | ||
2929 | #define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; }) | |
2930 | APPLY(s_commit_interval); | |
2931 | APPLY(s_stripe); | |
2932 | APPLY(s_max_batch_time); | |
2933 | APPLY(s_min_batch_time); | |
2934 | APPLY(s_want_extra_isize); | |
2935 | APPLY(s_inode_readahead_blks); | |
2936 | APPLY(s_max_dir_size_kb); | |
2937 | APPLY(s_li_wait_mult); | |
2938 | APPLY(s_resgid); | |
2939 | APPLY(s_resuid); | |
2940 | ||
2941 | #ifdef CONFIG_EXT4_DEBUG | |
2942 | APPLY(s_fc_debug_max_replay); | |
2943 | #endif | |
2944 | ||
2945 | ext4_apply_quota_options(fc, sb); | |
2946 | ||
2947 | if (ctx->spec & EXT4_SPEC_DUMMY_ENCRYPTION) | |
2948 | ret = ext4_set_test_dummy_encryption(sb, ctx->test_dummy_enc_arg); | |
2949 | ||
2950 | return ret; | |
2951 | } | |
2952 | ||
2953 | ||
2954 | static int ext4_validate_options(struct fs_context *fc) | |
2955 | { | |
ac27a0ec | 2956 | #ifdef CONFIG_QUOTA |
6e47a3cc | 2957 | struct ext4_fs_context *ctx = fc->fs_private; |
4c94bff9 | 2958 | char *usr_qf_name, *grp_qf_name; |
6e47a3cc LC |
2959 | |
2960 | usr_qf_name = ctx->s_qf_names[USRQUOTA]; | |
2961 | grp_qf_name = ctx->s_qf_names[GRPQUOTA]; | |
2962 | ||
33458eab | 2963 | if (usr_qf_name || grp_qf_name) { |
6e47a3cc LC |
2964 | if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name) |
2965 | ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA); | |
ac27a0ec | 2966 | |
6e47a3cc LC |
2967 | if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name) |
2968 | ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA); | |
ac27a0ec | 2969 | |
6e47a3cc LC |
2970 | if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) || |
2971 | ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) { | |
da812f61 | 2972 | ext4_msg(NULL, KERN_ERR, "old and new quota " |
6e47a3cc | 2973 | "format mixing"); |
da812f61 | 2974 | return -EINVAL; |
ac27a0ec | 2975 | } |
ac27a0ec DK |
2976 | } |
2977 | #endif | |
6e47a3cc | 2978 | return 1; |
ac27a0ec DK |
2979 | } |
2980 | ||
2adf6da8 TT |
2981 | static inline void ext4_show_quota_options(struct seq_file *seq, |
2982 | struct super_block *sb) | |
2983 | { | |
2984 | #if defined(CONFIG_QUOTA) | |
2985 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
33458eab | 2986 | char *usr_qf_name, *grp_qf_name; |
2adf6da8 TT |
2987 | |
2988 | if (sbi->s_jquota_fmt) { | |
2989 | char *fmtname = ""; | |
2990 | ||
2991 | switch (sbi->s_jquota_fmt) { | |
2992 | case QFMT_VFS_OLD: | |
2993 | fmtname = "vfsold"; | |
2994 | break; | |
2995 | case QFMT_VFS_V0: | |
2996 | fmtname = "vfsv0"; | |
2997 | break; | |
2998 | case QFMT_VFS_V1: | |
2999 | fmtname = "vfsv1"; | |
3000 | break; | |
3001 | } | |
3002 | seq_printf(seq, ",jqfmt=%s", fmtname); | |
3003 | } | |
3004 | ||
33458eab TT |
3005 | rcu_read_lock(); |
3006 | usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); | |
3007 | grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); | |
3008 | if (usr_qf_name) | |
3009 | seq_show_option(seq, "usrjquota", usr_qf_name); | |
3010 | if (grp_qf_name) | |
3011 | seq_show_option(seq, "grpjquota", grp_qf_name); | |
3012 | rcu_read_unlock(); | |
2adf6da8 TT |
3013 | #endif |
3014 | } | |
3015 | ||
5a916be1 TT |
3016 | static const char *token2str(int token) |
3017 | { | |
50df9fd5 | 3018 | const struct match_token *t; |
5a916be1 TT |
3019 | |
3020 | for (t = tokens; t->token != Opt_err; t++) | |
3021 | if (t->token == token && !strchr(t->pattern, '=')) | |
3022 | break; | |
3023 | return t->pattern; | |
3024 | } | |
3025 | ||
2adf6da8 TT |
3026 | /* |
3027 | * Show an option if | |
3028 | * - it's set to a non-default value OR | |
3029 | * - if the per-sb default is different from the global default | |
3030 | */ | |
66acdcf4 TT |
3031 | static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, |
3032 | int nodefs) | |
2adf6da8 | 3033 | { |
2adf6da8 TT |
3034 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
3035 | struct ext4_super_block *es = sbi->s_es; | |
68afa7e0 | 3036 | int def_errors, def_mount_opt = sbi->s_def_mount_opt; |
5a916be1 | 3037 | const struct mount_opts *m; |
66acdcf4 | 3038 | char sep = nodefs ? '\n' : ','; |
2adf6da8 | 3039 | |
66acdcf4 TT |
3040 | #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) |
3041 | #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) | |
2adf6da8 TT |
3042 | |
3043 | if (sbi->s_sb_block != 1) | |
5a916be1 TT |
3044 | SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); |
3045 | ||
3046 | for (m = ext4_mount_opts; m->token != Opt_err; m++) { | |
3047 | int want_set = m->flags & MOPT_SET; | |
3048 | if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || | |
9cb20f94 | 3049 | (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP) |
5a916be1 | 3050 | continue; |
68afa7e0 | 3051 | if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) |
5a916be1 TT |
3052 | continue; /* skip if same as the default */ |
3053 | if ((want_set && | |
3054 | (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || | |
3055 | (!want_set && (sbi->s_mount_opt & m->mount_opt))) | |
3056 | continue; /* select Opt_noFoo vs Opt_Foo */ | |
3057 | SEQ_OPTS_PRINT("%s", token2str(m->token)); | |
2adf6da8 | 3058 | } |
5a916be1 | 3059 | |
08cefc7a | 3060 | if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || |
5a916be1 | 3061 | le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) |
08cefc7a EB |
3062 | SEQ_OPTS_PRINT("resuid=%u", |
3063 | from_kuid_munged(&init_user_ns, sbi->s_resuid)); | |
3064 | if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || | |
5a916be1 | 3065 | le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) |
08cefc7a EB |
3066 | SEQ_OPTS_PRINT("resgid=%u", |
3067 | from_kgid_munged(&init_user_ns, sbi->s_resgid)); | |
66acdcf4 | 3068 | def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); |
5a916be1 TT |
3069 | if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) |
3070 | SEQ_OPTS_PUTS("errors=remount-ro"); | |
2adf6da8 | 3071 | if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) |
5a916be1 | 3072 | SEQ_OPTS_PUTS("errors=continue"); |
2adf6da8 | 3073 | if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) |
5a916be1 | 3074 | SEQ_OPTS_PUTS("errors=panic"); |
66acdcf4 | 3075 | if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) |
5a916be1 | 3076 | SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); |
66acdcf4 | 3077 | if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) |
5a916be1 | 3078 | SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); |
66acdcf4 | 3079 | if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) |
5a916be1 | 3080 | SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); |
357fdad0 | 3081 | if (sb->s_flags & SB_I_VERSION) |
5a916be1 | 3082 | SEQ_OPTS_PUTS("i_version"); |
66acdcf4 | 3083 | if (nodefs || sbi->s_stripe) |
5a916be1 | 3084 | SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); |
68afa7e0 TN |
3085 | if (nodefs || EXT4_MOUNT_DATA_FLAGS & |
3086 | (sbi->s_mount_opt ^ def_mount_opt)) { | |
5a916be1 TT |
3087 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) |
3088 | SEQ_OPTS_PUTS("data=journal"); | |
3089 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) | |
3090 | SEQ_OPTS_PUTS("data=ordered"); | |
3091 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) | |
3092 | SEQ_OPTS_PUTS("data=writeback"); | |
3093 | } | |
66acdcf4 TT |
3094 | if (nodefs || |
3095 | sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) | |
5a916be1 TT |
3096 | SEQ_OPTS_PRINT("inode_readahead_blks=%u", |
3097 | sbi->s_inode_readahead_blks); | |
2adf6da8 | 3098 | |
ceec0376 | 3099 | if (test_opt(sb, INIT_INODE_TABLE) && (nodefs || |
66acdcf4 | 3100 | (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) |
5a916be1 | 3101 | SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); |
df981d03 TT |
3102 | if (nodefs || sbi->s_max_dir_size_kb) |
3103 | SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); | |
7915a861 AN |
3104 | if (test_opt(sb, DATA_ERR_ABORT)) |
3105 | SEQ_OPTS_PUTS("data_err=abort"); | |
ed318a6c EB |
3106 | |
3107 | fscrypt_show_test_dummy_encryption(seq, sep, sb); | |
2adf6da8 | 3108 | |
4f74d15f EB |
3109 | if (sb->s_flags & SB_INLINECRYPT) |
3110 | SEQ_OPTS_PUTS("inlinecrypt"); | |
3111 | ||
9cb20f94 IW |
3112 | if (test_opt(sb, DAX_ALWAYS)) { |
3113 | if (IS_EXT2_SB(sb)) | |
3114 | SEQ_OPTS_PUTS("dax"); | |
3115 | else | |
3116 | SEQ_OPTS_PUTS("dax=always"); | |
3117 | } else if (test_opt2(sb, DAX_NEVER)) { | |
3118 | SEQ_OPTS_PUTS("dax=never"); | |
3119 | } else if (test_opt2(sb, DAX_INODE)) { | |
3120 | SEQ_OPTS_PUTS("dax=inode"); | |
3121 | } | |
2adf6da8 | 3122 | ext4_show_quota_options(seq, sb); |
2adf6da8 TT |
3123 | return 0; |
3124 | } | |
3125 | ||
66acdcf4 TT |
3126 | static int ext4_show_options(struct seq_file *seq, struct dentry *root) |
3127 | { | |
3128 | return _ext4_show_options(seq, root->d_sb, 0); | |
3129 | } | |
3130 | ||
ebd173be | 3131 | int ext4_seq_options_show(struct seq_file *seq, void *offset) |
66acdcf4 TT |
3132 | { |
3133 | struct super_block *sb = seq->private; | |
3134 | int rc; | |
3135 | ||
bc98a42c | 3136 | seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw"); |
66acdcf4 TT |
3137 | rc = _ext4_show_options(seq, sb, 1); |
3138 | seq_puts(seq, "\n"); | |
3139 | return rc; | |
3140 | } | |
3141 | ||
617ba13b | 3142 | static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, |
ac27a0ec DK |
3143 | int read_only) |
3144 | { | |
617ba13b | 3145 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
c89128a0 | 3146 | int err = 0; |
ac27a0ec | 3147 | |
617ba13b | 3148 | if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { |
b31e1552 ES |
3149 | ext4_msg(sb, KERN_ERR, "revision level too high, " |
3150 | "forcing read-only mode"); | |
c89128a0 | 3151 | err = -EROFS; |
5adaccac | 3152 | goto done; |
ac27a0ec DK |
3153 | } |
3154 | if (read_only) | |
281b5995 | 3155 | goto done; |
617ba13b | 3156 | if (!(sbi->s_mount_state & EXT4_VALID_FS)) |
b31e1552 ES |
3157 | ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " |
3158 | "running e2fsck is recommended"); | |
c8b459f4 | 3159 | else if (sbi->s_mount_state & EXT4_ERROR_FS) |
b31e1552 ES |
3160 | ext4_msg(sb, KERN_WARNING, |
3161 | "warning: mounting fs with errors, " | |
3162 | "running e2fsck is recommended"); | |
ed3ce80a | 3163 | else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && |
ac27a0ec DK |
3164 | le16_to_cpu(es->s_mnt_count) >= |
3165 | (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) | |
b31e1552 ES |
3166 | ext4_msg(sb, KERN_WARNING, |
3167 | "warning: maximal mount count reached, " | |
3168 | "running e2fsck is recommended"); | |
ac27a0ec | 3169 | else if (le32_to_cpu(es->s_checkinterval) && |
6a0678a7 AB |
3170 | (ext4_get_tstamp(es, s_lastcheck) + |
3171 | le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds())) | |
b31e1552 ES |
3172 | ext4_msg(sb, KERN_WARNING, |
3173 | "warning: checktime reached, " | |
3174 | "running e2fsck is recommended"); | |
0b8e58a1 | 3175 | if (!sbi->s_journal) |
0390131b | 3176 | es->s_state &= cpu_to_le16(~EXT4_VALID_FS); |
ac27a0ec | 3177 | if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) |
617ba13b | 3178 | es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); |
e8546d06 | 3179 | le16_add_cpu(&es->s_mnt_count, 1); |
6a0678a7 | 3180 | ext4_update_tstamp(es, s_mtime); |
02f310fc | 3181 | if (sbi->s_journal) { |
e2b911c5 | 3182 | ext4_set_feature_journal_needs_recovery(sb); |
02f310fc JK |
3183 | if (ext4_has_feature_orphan_file(sb)) |
3184 | ext4_set_feature_orphan_present(sb); | |
3185 | } | |
ac27a0ec | 3186 | |
4392fbc4 | 3187 | err = ext4_commit_super(sb); |
281b5995 | 3188 | done: |
ac27a0ec | 3189 | if (test_opt(sb, DEBUG)) |
a9df9a49 | 3190 | printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " |
a2595b8a | 3191 | "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", |
ac27a0ec DK |
3192 | sb->s_blocksize, |
3193 | sbi->s_groups_count, | |
617ba13b MC |
3194 | EXT4_BLOCKS_PER_GROUP(sb), |
3195 | EXT4_INODES_PER_GROUP(sb), | |
a2595b8a | 3196 | sbi->s_mount_opt, sbi->s_mount_opt2); |
ac27a0ec | 3197 | |
7abc52c2 | 3198 | cleancache_init_fs(sb); |
c89128a0 | 3199 | return err; |
ac27a0ec DK |
3200 | } |
3201 | ||
117fff10 TT |
3202 | int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) |
3203 | { | |
3204 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
7c990728 | 3205 | struct flex_groups **old_groups, **new_groups; |
37b0b6b8 | 3206 | int size, i, j; |
117fff10 TT |
3207 | |
3208 | if (!sbi->s_log_groups_per_flex) | |
3209 | return 0; | |
3210 | ||
3211 | size = ext4_flex_group(sbi, ngroup - 1) + 1; | |
3212 | if (size <= sbi->s_flex_groups_allocated) | |
3213 | return 0; | |
3214 | ||
7c990728 SJS |
3215 | new_groups = kvzalloc(roundup_pow_of_two(size * |
3216 | sizeof(*sbi->s_flex_groups)), GFP_KERNEL); | |
117fff10 | 3217 | if (!new_groups) { |
7c990728 SJS |
3218 | ext4_msg(sb, KERN_ERR, |
3219 | "not enough memory for %d flex group pointers", size); | |
117fff10 TT |
3220 | return -ENOMEM; |
3221 | } | |
7c990728 SJS |
3222 | for (i = sbi->s_flex_groups_allocated; i < size; i++) { |
3223 | new_groups[i] = kvzalloc(roundup_pow_of_two( | |
3224 | sizeof(struct flex_groups)), | |
3225 | GFP_KERNEL); | |
3226 | if (!new_groups[i]) { | |
37b0b6b8 DC |
3227 | for (j = sbi->s_flex_groups_allocated; j < i; j++) |
3228 | kvfree(new_groups[j]); | |
7c990728 SJS |
3229 | kvfree(new_groups); |
3230 | ext4_msg(sb, KERN_ERR, | |
3231 | "not enough memory for %d flex groups", size); | |
3232 | return -ENOMEM; | |
3233 | } | |
117fff10 | 3234 | } |
7c990728 SJS |
3235 | rcu_read_lock(); |
3236 | old_groups = rcu_dereference(sbi->s_flex_groups); | |
3237 | if (old_groups) | |
3238 | memcpy(new_groups, old_groups, | |
3239 | (sbi->s_flex_groups_allocated * | |
3240 | sizeof(struct flex_groups *))); | |
3241 | rcu_read_unlock(); | |
3242 | rcu_assign_pointer(sbi->s_flex_groups, new_groups); | |
3243 | sbi->s_flex_groups_allocated = size; | |
3244 | if (old_groups) | |
3245 | ext4_kvfree_array_rcu(old_groups); | |
117fff10 TT |
3246 | return 0; |
3247 | } | |
3248 | ||
772cb7c8 JS |
3249 | static int ext4_fill_flex_info(struct super_block *sb) |
3250 | { | |
3251 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
3252 | struct ext4_group_desc *gdp = NULL; | |
7c990728 | 3253 | struct flex_groups *fg; |
772cb7c8 | 3254 | ext4_group_t flex_group; |
117fff10 | 3255 | int i, err; |
772cb7c8 | 3256 | |
503358ae | 3257 | sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; |
d50f2ab6 | 3258 | if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { |
772cb7c8 JS |
3259 | sbi->s_log_groups_per_flex = 0; |
3260 | return 1; | |
3261 | } | |
3262 | ||
117fff10 TT |
3263 | err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); |
3264 | if (err) | |
9933fc0a | 3265 | goto failed; |
772cb7c8 | 3266 | |
772cb7c8 | 3267 | for (i = 0; i < sbi->s_groups_count; i++) { |
88b6edd1 | 3268 | gdp = ext4_get_group_desc(sb, i, NULL); |
772cb7c8 JS |
3269 | |
3270 | flex_group = ext4_flex_group(sbi, i); | |
7c990728 SJS |
3271 | fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group); |
3272 | atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes); | |
90ba983f | 3273 | atomic64_add(ext4_free_group_clusters(sb, gdp), |
7c990728 SJS |
3274 | &fg->free_clusters); |
3275 | atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs); | |
772cb7c8 JS |
3276 | } |
3277 | ||
3278 | return 1; | |
3279 | failed: | |
3280 | return 0; | |
3281 | } | |
3282 | ||
e2b911c5 | 3283 | static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, |
feb0ab32 | 3284 | struct ext4_group_desc *gdp) |
717d50e4 | 3285 | { |
b47820ed | 3286 | int offset = offsetof(struct ext4_group_desc, bg_checksum); |
717d50e4 | 3287 | __u16 crc = 0; |
feb0ab32 | 3288 | __le32 le_group = cpu_to_le32(block_group); |
e2b911c5 | 3289 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
717d50e4 | 3290 | |
9aa5d32b | 3291 | if (ext4_has_metadata_csum(sbi->s_sb)) { |
feb0ab32 | 3292 | /* Use new metadata_csum algorithm */ |
feb0ab32 | 3293 | __u32 csum32; |
b47820ed | 3294 | __u16 dummy_csum = 0; |
feb0ab32 | 3295 | |
feb0ab32 DW |
3296 | csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, |
3297 | sizeof(le_group)); | |
b47820ed DJ |
3298 | csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset); |
3299 | csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum, | |
3300 | sizeof(dummy_csum)); | |
3301 | offset += sizeof(dummy_csum); | |
3302 | if (offset < sbi->s_desc_size) | |
3303 | csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset, | |
3304 | sbi->s_desc_size - offset); | |
feb0ab32 DW |
3305 | |
3306 | crc = csum32 & 0xFFFF; | |
3307 | goto out; | |
717d50e4 AD |
3308 | } |
3309 | ||
feb0ab32 | 3310 | /* old crc16 code */ |
e2b911c5 | 3311 | if (!ext4_has_feature_gdt_csum(sb)) |
813d32f9 DW |
3312 | return 0; |
3313 | ||
feb0ab32 DW |
3314 | crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); |
3315 | crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); | |
3316 | crc = crc16(crc, (__u8 *)gdp, offset); | |
3317 | offset += sizeof(gdp->bg_checksum); /* skip checksum */ | |
3318 | /* for checksum of struct ext4_group_desc do the rest...*/ | |
e2b911c5 | 3319 | if (ext4_has_feature_64bit(sb) && |
feb0ab32 DW |
3320 | offset < le16_to_cpu(sbi->s_es->s_desc_size)) |
3321 | crc = crc16(crc, (__u8 *)gdp + offset, | |
3322 | le16_to_cpu(sbi->s_es->s_desc_size) - | |
3323 | offset); | |
3324 | ||
3325 | out: | |
717d50e4 AD |
3326 | return cpu_to_le16(crc); |
3327 | } | |
3328 | ||
feb0ab32 | 3329 | int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, |
717d50e4 AD |
3330 | struct ext4_group_desc *gdp) |
3331 | { | |
feb0ab32 | 3332 | if (ext4_has_group_desc_csum(sb) && |
e2b911c5 | 3333 | (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) |
717d50e4 AD |
3334 | return 0; |
3335 | ||
3336 | return 1; | |
3337 | } | |
3338 | ||
feb0ab32 DW |
3339 | void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, |
3340 | struct ext4_group_desc *gdp) | |
3341 | { | |
3342 | if (!ext4_has_group_desc_csum(sb)) | |
3343 | return; | |
e2b911c5 | 3344 | gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); |
feb0ab32 DW |
3345 | } |
3346 | ||
ac27a0ec | 3347 | /* Called at mount-time, super-block is locked */ |
bfff6873 | 3348 | static int ext4_check_descriptors(struct super_block *sb, |
829fa70d | 3349 | ext4_fsblk_t sb_block, |
bfff6873 | 3350 | ext4_group_t *first_not_zeroed) |
ac27a0ec | 3351 | { |
617ba13b MC |
3352 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
3353 | ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); | |
3354 | ext4_fsblk_t last_block; | |
44de022c | 3355 | ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0); |
bd81d8ee LV |
3356 | ext4_fsblk_t block_bitmap; |
3357 | ext4_fsblk_t inode_bitmap; | |
3358 | ext4_fsblk_t inode_table; | |
ce421581 | 3359 | int flexbg_flag = 0; |
bfff6873 | 3360 | ext4_group_t i, grp = sbi->s_groups_count; |
ac27a0ec | 3361 | |
e2b911c5 | 3362 | if (ext4_has_feature_flex_bg(sb)) |
ce421581 JS |
3363 | flexbg_flag = 1; |
3364 | ||
af5bc92d | 3365 | ext4_debug("Checking group descriptors"); |
ac27a0ec | 3366 | |
197cd65a AM |
3367 | for (i = 0; i < sbi->s_groups_count; i++) { |
3368 | struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); | |
3369 | ||
ce421581 | 3370 | if (i == sbi->s_groups_count - 1 || flexbg_flag) |
bd81d8ee | 3371 | last_block = ext4_blocks_count(sbi->s_es) - 1; |
ac27a0ec DK |
3372 | else |
3373 | last_block = first_block + | |
617ba13b | 3374 | (EXT4_BLOCKS_PER_GROUP(sb) - 1); |
ac27a0ec | 3375 | |
bfff6873 LC |
3376 | if ((grp == sbi->s_groups_count) && |
3377 | !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) | |
3378 | grp = i; | |
3379 | ||
8fadc143 | 3380 | block_bitmap = ext4_block_bitmap(sb, gdp); |
829fa70d TT |
3381 | if (block_bitmap == sb_block) { |
3382 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " | |
3383 | "Block bitmap for group %u overlaps " | |
3384 | "superblock", i); | |
18db4b4e TT |
3385 | if (!sb_rdonly(sb)) |
3386 | return 0; | |
829fa70d | 3387 | } |
77260807 TT |
3388 | if (block_bitmap >= sb_block + 1 && |
3389 | block_bitmap <= last_bg_block) { | |
3390 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " | |
3391 | "Block bitmap for group %u overlaps " | |
3392 | "block group descriptors", i); | |
3393 | if (!sb_rdonly(sb)) | |
3394 | return 0; | |
3395 | } | |
2b2d6d01 | 3396 | if (block_bitmap < first_block || block_bitmap > last_block) { |
b31e1552 | 3397 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 3398 | "Block bitmap for group %u not in group " |
b31e1552 | 3399 | "(block %llu)!", i, block_bitmap); |
ac27a0ec DK |
3400 | return 0; |
3401 | } | |
8fadc143 | 3402 | inode_bitmap = ext4_inode_bitmap(sb, gdp); |
829fa70d TT |
3403 | if (inode_bitmap == sb_block) { |
3404 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " | |
3405 | "Inode bitmap for group %u overlaps " | |
3406 | "superblock", i); | |
18db4b4e TT |
3407 | if (!sb_rdonly(sb)) |
3408 | return 0; | |
829fa70d | 3409 | } |
77260807 TT |
3410 | if (inode_bitmap >= sb_block + 1 && |
3411 | inode_bitmap <= last_bg_block) { | |
3412 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " | |
3413 | "Inode bitmap for group %u overlaps " | |
3414 | "block group descriptors", i); | |
3415 | if (!sb_rdonly(sb)) | |
3416 | return 0; | |
3417 | } | |
2b2d6d01 | 3418 | if (inode_bitmap < first_block || inode_bitmap > last_block) { |
b31e1552 | 3419 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 3420 | "Inode bitmap for group %u not in group " |
b31e1552 | 3421 | "(block %llu)!", i, inode_bitmap); |
ac27a0ec DK |
3422 | return 0; |
3423 | } | |
8fadc143 | 3424 | inode_table = ext4_inode_table(sb, gdp); |
829fa70d TT |
3425 | if (inode_table == sb_block) { |
3426 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " | |
3427 | "Inode table for group %u overlaps " | |
3428 | "superblock", i); | |
18db4b4e TT |
3429 | if (!sb_rdonly(sb)) |
3430 | return 0; | |
829fa70d | 3431 | } |
77260807 TT |
3432 | if (inode_table >= sb_block + 1 && |
3433 | inode_table <= last_bg_block) { | |
3434 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " | |
3435 | "Inode table for group %u overlaps " | |
3436 | "block group descriptors", i); | |
3437 | if (!sb_rdonly(sb)) | |
3438 | return 0; | |
3439 | } | |
bd81d8ee | 3440 | if (inode_table < first_block || |
2b2d6d01 | 3441 | inode_table + sbi->s_itb_per_group - 1 > last_block) { |
b31e1552 | 3442 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
a9df9a49 | 3443 | "Inode table for group %u not in group " |
b31e1552 | 3444 | "(block %llu)!", i, inode_table); |
ac27a0ec DK |
3445 | return 0; |
3446 | } | |
955ce5f5 | 3447 | ext4_lock_group(sb, i); |
feb0ab32 | 3448 | if (!ext4_group_desc_csum_verify(sb, i, gdp)) { |
b31e1552 ES |
3449 | ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " |
3450 | "Checksum for group %u failed (%u!=%u)", | |
e2b911c5 | 3451 | i, le16_to_cpu(ext4_group_desc_csum(sb, i, |
b31e1552 | 3452 | gdp)), le16_to_cpu(gdp->bg_checksum)); |
bc98a42c | 3453 | if (!sb_rdonly(sb)) { |
955ce5f5 | 3454 | ext4_unlock_group(sb, i); |
8a266467 | 3455 | return 0; |
7ee1ec4c | 3456 | } |
717d50e4 | 3457 | } |
955ce5f5 | 3458 | ext4_unlock_group(sb, i); |
ce421581 JS |
3459 | if (!flexbg_flag) |
3460 | first_block += EXT4_BLOCKS_PER_GROUP(sb); | |
ac27a0ec | 3461 | } |
bfff6873 LC |
3462 | if (NULL != first_not_zeroed) |
3463 | *first_not_zeroed = grp; | |
ac27a0ec DK |
3464 | return 1; |
3465 | } | |
3466 | ||
cd2291a4 ES |
3467 | /* |
3468 | * Maximal extent format file size. | |
3469 | * Resulting logical blkno at s_maxbytes must fit in our on-disk | |
3470 | * extent format containers, within a sector_t, and within i_blocks | |
3471 | * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, | |
3472 | * so that won't be a limiting factor. | |
3473 | * | |
f17722f9 LC |
3474 | * However there is other limiting factor. We do store extents in the form |
3475 | * of starting block and length, hence the resulting length of the extent | |
3476 | * covering maximum file size must fit into on-disk format containers as | |
3477 | * well. Given that length is always by 1 unit bigger than max unit (because | |
3478 | * we count 0 as well) we have to lower the s_maxbytes by one fs block. | |
3479 | * | |
cd2291a4 ES |
3480 | * Note, this does *not* consider any metadata overhead for vfs i_blocks. |
3481 | */ | |
f287a1a5 | 3482 | static loff_t ext4_max_size(int blkbits, int has_huge_files) |
cd2291a4 ES |
3483 | { |
3484 | loff_t res; | |
3485 | loff_t upper_limit = MAX_LFS_FILESIZE; | |
3486 | ||
72deb455 CH |
3487 | BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64)); |
3488 | ||
3489 | if (!has_huge_files) { | |
cd2291a4 ES |
3490 | upper_limit = (1LL << 32) - 1; |
3491 | ||
3492 | /* total blocks in file system block size */ | |
3493 | upper_limit >>= (blkbits - 9); | |
3494 | upper_limit <<= blkbits; | |
3495 | } | |
3496 | ||
f17722f9 LC |
3497 | /* |
3498 | * 32-bit extent-start container, ee_block. We lower the maxbytes | |
3499 | * by one fs block, so ee_len can cover the extent of maximum file | |
3500 | * size | |
3501 | */ | |
3502 | res = (1LL << 32) - 1; | |
cd2291a4 | 3503 | res <<= blkbits; |
cd2291a4 ES |
3504 | |
3505 | /* Sanity check against vm- & vfs- imposed limits */ | |
3506 | if (res > upper_limit) | |
3507 | res = upper_limit; | |
3508 | ||
3509 | return res; | |
3510 | } | |
ac27a0ec | 3511 | |
ac27a0ec | 3512 | /* |
cd2291a4 | 3513 | * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect |
0fc1b451 AK |
3514 | * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. |
3515 | * We need to be 1 filesystem block less than the 2^48 sector limit. | |
ac27a0ec | 3516 | */ |
f287a1a5 | 3517 | static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) |
ac27a0ec | 3518 | { |
75ca6ad4 | 3519 | unsigned long long upper_limit, res = EXT4_NDIR_BLOCKS; |
0fc1b451 | 3520 | int meta_blocks; |
75ca6ad4 RH |
3521 | |
3522 | /* | |
3523 | * This is calculated to be the largest file size for a dense, block | |
0b8e58a1 AD |
3524 | * mapped file such that the file's total number of 512-byte sectors, |
3525 | * including data and all indirect blocks, does not exceed (2^48 - 1). | |
3526 | * | |
3527 | * __u32 i_blocks_lo and _u16 i_blocks_high represent the total | |
3528 | * number of 512-byte sectors of the file. | |
0fc1b451 | 3529 | */ |
72deb455 | 3530 | if (!has_huge_files) { |
0fc1b451 | 3531 | /* |
72deb455 CH |
3532 | * !has_huge_files or implies that the inode i_block field |
3533 | * represents total file blocks in 2^32 512-byte sectors == | |
3534 | * size of vfs inode i_blocks * 8 | |
0fc1b451 AK |
3535 | */ |
3536 | upper_limit = (1LL << 32) - 1; | |
3537 | ||
3538 | /* total blocks in file system block size */ | |
3539 | upper_limit >>= (bits - 9); | |
3540 | ||
3541 | } else { | |
8180a562 AK |
3542 | /* |
3543 | * We use 48 bit ext4_inode i_blocks | |
3544 | * With EXT4_HUGE_FILE_FL set the i_blocks | |
3545 | * represent total number of blocks in | |
3546 | * file system block size | |
3547 | */ | |
0fc1b451 AK |
3548 | upper_limit = (1LL << 48) - 1; |
3549 | ||
0fc1b451 AK |
3550 | } |
3551 | ||
3552 | /* indirect blocks */ | |
3553 | meta_blocks = 1; | |
3554 | /* double indirect blocks */ | |
3555 | meta_blocks += 1 + (1LL << (bits-2)); | |
3556 | /* tripple indirect blocks */ | |
3557 | meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); | |
3558 | ||
3559 | upper_limit -= meta_blocks; | |
3560 | upper_limit <<= bits; | |
ac27a0ec DK |
3561 | |
3562 | res += 1LL << (bits-2); | |
3563 | res += 1LL << (2*(bits-2)); | |
3564 | res += 1LL << (3*(bits-2)); | |
3565 | res <<= bits; | |
3566 | if (res > upper_limit) | |
3567 | res = upper_limit; | |
0fc1b451 AK |
3568 | |
3569 | if (res > MAX_LFS_FILESIZE) | |
3570 | res = MAX_LFS_FILESIZE; | |
3571 | ||
75ca6ad4 | 3572 | return (loff_t)res; |
ac27a0ec DK |
3573 | } |
3574 | ||
617ba13b | 3575 | static ext4_fsblk_t descriptor_loc(struct super_block *sb, |
0b8e58a1 | 3576 | ext4_fsblk_t logical_sb_block, int nr) |
ac27a0ec | 3577 | { |
617ba13b | 3578 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
fd2d4291 | 3579 | ext4_group_t bg, first_meta_bg; |
ac27a0ec DK |
3580 | int has_super = 0; |
3581 | ||
3582 | first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); | |
3583 | ||
e2b911c5 | 3584 | if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) |
70bbb3e0 | 3585 | return logical_sb_block + nr + 1; |
ac27a0ec | 3586 | bg = sbi->s_desc_per_block * nr; |
617ba13b | 3587 | if (ext4_bg_has_super(sb, bg)) |
ac27a0ec | 3588 | has_super = 1; |
0b8e58a1 | 3589 | |
bd63f6b0 DW |
3590 | /* |
3591 | * If we have a meta_bg fs with 1k blocks, group 0's GDT is at | |
3592 | * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled | |
3593 | * on modern mke2fs or blksize > 1k on older mke2fs) then we must | |
3594 | * compensate. | |
3595 | */ | |
3596 | if (sb->s_blocksize == 1024 && nr == 0 && | |
49598e04 | 3597 | le32_to_cpu(sbi->s_es->s_first_data_block) == 0) |
bd63f6b0 DW |
3598 | has_super++; |
3599 | ||
617ba13b | 3600 | return (has_super + ext4_group_first_block_no(sb, bg)); |
ac27a0ec DK |
3601 | } |
3602 | ||
c9de560d AT |
3603 | /** |
3604 | * ext4_get_stripe_size: Get the stripe size. | |
3605 | * @sbi: In memory super block info | |
3606 | * | |
3607 | * If we have specified it via mount option, then | |
3608 | * use the mount option value. If the value specified at mount time is | |
3609 | * greater than the blocks per group use the super block value. | |
3610 | * If the super block value is greater than blocks per group return 0. | |
3611 | * Allocator needs it be less than blocks per group. | |
3612 | * | |
3613 | */ | |
3614 | static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) | |
3615 | { | |
3616 | unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); | |
3617 | unsigned long stripe_width = | |
3618 | le32_to_cpu(sbi->s_es->s_raid_stripe_width); | |
3eb08658 | 3619 | int ret; |
c9de560d AT |
3620 | |
3621 | if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) | |
3eb08658 | 3622 | ret = sbi->s_stripe; |
5469d7c3 | 3623 | else if (stripe_width && stripe_width <= sbi->s_blocks_per_group) |
3eb08658 | 3624 | ret = stripe_width; |
5469d7c3 | 3625 | else if (stride && stride <= sbi->s_blocks_per_group) |
3eb08658 DE |
3626 | ret = stride; |
3627 | else | |
3628 | ret = 0; | |
c9de560d | 3629 | |
3eb08658 DE |
3630 | /* |
3631 | * If the stripe width is 1, this makes no sense and | |
3632 | * we set it to 0 to turn off stripe handling code. | |
3633 | */ | |
3634 | if (ret <= 1) | |
3635 | ret = 0; | |
c9de560d | 3636 | |
3eb08658 | 3637 | return ret; |
c9de560d | 3638 | } |
ac27a0ec | 3639 | |
a13fb1a4 ES |
3640 | /* |
3641 | * Check whether this filesystem can be mounted based on | |
3642 | * the features present and the RDONLY/RDWR mount requested. | |
3643 | * Returns 1 if this filesystem can be mounted as requested, | |
3644 | * 0 if it cannot be. | |
3645 | */ | |
25c6d98f | 3646 | int ext4_feature_set_ok(struct super_block *sb, int readonly) |
a13fb1a4 | 3647 | { |
e2b911c5 | 3648 | if (ext4_has_unknown_ext4_incompat_features(sb)) { |
a13fb1a4 ES |
3649 | ext4_msg(sb, KERN_ERR, |
3650 | "Couldn't mount because of " | |
3651 | "unsupported optional features (%x)", | |
3652 | (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & | |
3653 | ~EXT4_FEATURE_INCOMPAT_SUPP)); | |
3654 | return 0; | |
3655 | } | |
3656 | ||
c83ad55e GKB |
3657 | #ifndef CONFIG_UNICODE |
3658 | if (ext4_has_feature_casefold(sb)) { | |
3659 | ext4_msg(sb, KERN_ERR, | |
3660 | "Filesystem with casefold feature cannot be " | |
3661 | "mounted without CONFIG_UNICODE"); | |
3662 | return 0; | |
3663 | } | |
3664 | #endif | |
3665 | ||
a13fb1a4 ES |
3666 | if (readonly) |
3667 | return 1; | |
3668 | ||
e2b911c5 | 3669 | if (ext4_has_feature_readonly(sb)) { |
2cb5cc8b | 3670 | ext4_msg(sb, KERN_INFO, "filesystem is read-only"); |
1751e8a6 | 3671 | sb->s_flags |= SB_RDONLY; |
2cb5cc8b DW |
3672 | return 1; |
3673 | } | |
3674 | ||
a13fb1a4 | 3675 | /* Check that feature set is OK for a read-write mount */ |
e2b911c5 | 3676 | if (ext4_has_unknown_ext4_ro_compat_features(sb)) { |
a13fb1a4 ES |
3677 | ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " |
3678 | "unsupported optional features (%x)", | |
3679 | (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & | |
3680 | ~EXT4_FEATURE_RO_COMPAT_SUPP)); | |
3681 | return 0; | |
3682 | } | |
e2b911c5 | 3683 | if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { |
bab08ab9 TT |
3684 | ext4_msg(sb, KERN_ERR, |
3685 | "Can't support bigalloc feature without " | |
3686 | "extents feature\n"); | |
3687 | return 0; | |
3688 | } | |
7c319d32 | 3689 | |
9db176bc | 3690 | #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2) |
d65d87a0 TT |
3691 | if (!readonly && (ext4_has_feature_quota(sb) || |
3692 | ext4_has_feature_project(sb))) { | |
7c319d32 | 3693 | ext4_msg(sb, KERN_ERR, |
d65d87a0 | 3694 | "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2"); |
689c958c LX |
3695 | return 0; |
3696 | } | |
7c319d32 | 3697 | #endif /* CONFIG_QUOTA */ |
a13fb1a4 ES |
3698 | return 1; |
3699 | } | |
3700 | ||
66e61a9e TT |
3701 | /* |
3702 | * This function is called once a day if we have errors logged | |
3703 | * on the file system | |
3704 | */ | |
235699a8 | 3705 | static void print_daily_error_info(struct timer_list *t) |
66e61a9e | 3706 | { |
235699a8 KC |
3707 | struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report); |
3708 | struct super_block *sb = sbi->s_sb; | |
3709 | struct ext4_super_block *es = sbi->s_es; | |
66e61a9e TT |
3710 | |
3711 | if (es->s_error_count) | |
ae0f78de TT |
3712 | /* fsck newer than v1.41.13 is needed to clean this condition. */ |
3713 | ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", | |
66e61a9e TT |
3714 | le32_to_cpu(es->s_error_count)); |
3715 | if (es->s_first_error_time) { | |
6a0678a7 AB |
3716 | printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d", |
3717 | sb->s_id, | |
3718 | ext4_get_tstamp(es, s_first_error_time), | |
66e61a9e TT |
3719 | (int) sizeof(es->s_first_error_func), |
3720 | es->s_first_error_func, | |
3721 | le32_to_cpu(es->s_first_error_line)); | |
3722 | if (es->s_first_error_ino) | |
651e1c3b | 3723 | printk(KERN_CONT ": inode %u", |
66e61a9e TT |
3724 | le32_to_cpu(es->s_first_error_ino)); |
3725 | if (es->s_first_error_block) | |
651e1c3b | 3726 | printk(KERN_CONT ": block %llu", (unsigned long long) |
66e61a9e | 3727 | le64_to_cpu(es->s_first_error_block)); |
651e1c3b | 3728 | printk(KERN_CONT "\n"); |
66e61a9e TT |
3729 | } |
3730 | if (es->s_last_error_time) { | |
6a0678a7 AB |
3731 | printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d", |
3732 | sb->s_id, | |
3733 | ext4_get_tstamp(es, s_last_error_time), | |
66e61a9e TT |
3734 | (int) sizeof(es->s_last_error_func), |
3735 | es->s_last_error_func, | |
3736 | le32_to_cpu(es->s_last_error_line)); | |
3737 | if (es->s_last_error_ino) | |
651e1c3b | 3738 | printk(KERN_CONT ": inode %u", |
66e61a9e TT |
3739 | le32_to_cpu(es->s_last_error_ino)); |
3740 | if (es->s_last_error_block) | |
651e1c3b | 3741 | printk(KERN_CONT ": block %llu", (unsigned long long) |
66e61a9e | 3742 | le64_to_cpu(es->s_last_error_block)); |
651e1c3b | 3743 | printk(KERN_CONT "\n"); |
66e61a9e TT |
3744 | } |
3745 | mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ | |
3746 | } | |
3747 | ||
bfff6873 LC |
3748 | /* Find next suitable group and run ext4_init_inode_table */ |
3749 | static int ext4_run_li_request(struct ext4_li_request *elr) | |
3750 | { | |
3751 | struct ext4_group_desc *gdp = NULL; | |
3d392b26 TT |
3752 | struct super_block *sb = elr->lr_super; |
3753 | ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; | |
3754 | ext4_group_t group = elr->lr_next_group; | |
3d392b26 | 3755 | unsigned int prefetch_ios = 0; |
bfff6873 | 3756 | int ret = 0; |
39fec688 | 3757 | u64 start_time; |
bfff6873 | 3758 | |
3d392b26 TT |
3759 | if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) { |
3760 | elr->lr_next_group = ext4_mb_prefetch(sb, group, | |
3761 | EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios); | |
3762 | if (prefetch_ios) | |
3763 | ext4_mb_prefetch_fini(sb, elr->lr_next_group, | |
3764 | prefetch_ios); | |
3765 | trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group, | |
3766 | prefetch_ios); | |
3767 | if (group >= elr->lr_next_group) { | |
3768 | ret = 1; | |
3769 | if (elr->lr_first_not_zeroed != ngroups && | |
3770 | !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) { | |
3771 | elr->lr_next_group = elr->lr_first_not_zeroed; | |
3772 | elr->lr_mode = EXT4_LI_MODE_ITABLE; | |
3773 | ret = 0; | |
3774 | } | |
3775 | } | |
3776 | return ret; | |
3777 | } | |
bfff6873 | 3778 | |
3d392b26 | 3779 | for (; group < ngroups; group++) { |
bfff6873 LC |
3780 | gdp = ext4_get_group_desc(sb, group, NULL); |
3781 | if (!gdp) { | |
3782 | ret = 1; | |
3783 | break; | |
3784 | } | |
3785 | ||
3786 | if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) | |
3787 | break; | |
3788 | } | |
3789 | ||
7f511862 | 3790 | if (group >= ngroups) |
bfff6873 LC |
3791 | ret = 1; |
3792 | ||
3793 | if (!ret) { | |
39fec688 | 3794 | start_time = ktime_get_real_ns(); |
bfff6873 LC |
3795 | ret = ext4_init_inode_table(sb, group, |
3796 | elr->lr_timeout ? 0 : 1); | |
3d392b26 | 3797 | trace_ext4_lazy_itable_init(sb, group); |
bfff6873 | 3798 | if (elr->lr_timeout == 0) { |
39fec688 SX |
3799 | elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) * |
3800 | EXT4_SB(elr->lr_super)->s_li_wait_mult); | |
bfff6873 LC |
3801 | } |
3802 | elr->lr_next_sched = jiffies + elr->lr_timeout; | |
3803 | elr->lr_next_group = group + 1; | |
3804 | } | |
bfff6873 LC |
3805 | return ret; |
3806 | } | |
3807 | ||
3808 | /* | |
3809 | * Remove lr_request from the list_request and free the | |
4ed5c033 | 3810 | * request structure. Should be called with li_list_mtx held |
bfff6873 LC |
3811 | */ |
3812 | static void ext4_remove_li_request(struct ext4_li_request *elr) | |
3813 | { | |
bfff6873 LC |
3814 | if (!elr) |
3815 | return; | |
3816 | ||
bfff6873 | 3817 | list_del(&elr->lr_request); |
3d392b26 | 3818 | EXT4_SB(elr->lr_super)->s_li_request = NULL; |
bfff6873 LC |
3819 | kfree(elr); |
3820 | } | |
3821 | ||
3822 | static void ext4_unregister_li_request(struct super_block *sb) | |
3823 | { | |
1bb933fb LC |
3824 | mutex_lock(&ext4_li_mtx); |
3825 | if (!ext4_li_info) { | |
3826 | mutex_unlock(&ext4_li_mtx); | |
bfff6873 | 3827 | return; |
1bb933fb | 3828 | } |
bfff6873 LC |
3829 | |
3830 | mutex_lock(&ext4_li_info->li_list_mtx); | |
1bb933fb | 3831 | ext4_remove_li_request(EXT4_SB(sb)->s_li_request); |
bfff6873 | 3832 | mutex_unlock(&ext4_li_info->li_list_mtx); |
1bb933fb | 3833 | mutex_unlock(&ext4_li_mtx); |
bfff6873 LC |
3834 | } |
3835 | ||
8f1f7453 ES |
3836 | static struct task_struct *ext4_lazyinit_task; |
3837 | ||
bfff6873 LC |
3838 | /* |
3839 | * This is the function where ext4lazyinit thread lives. It walks | |
3840 | * through the request list searching for next scheduled filesystem. | |
3841 | * When such a fs is found, run the lazy initialization request | |
3842 | * (ext4_rn_li_request) and keep track of the time spend in this | |
3843 | * function. Based on that time we compute next schedule time of | |
3844 | * the request. When walking through the list is complete, compute | |
3845 | * next waking time and put itself into sleep. | |
3846 | */ | |
3847 | static int ext4_lazyinit_thread(void *arg) | |
3848 | { | |
3849 | struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; | |
3850 | struct list_head *pos, *n; | |
3851 | struct ext4_li_request *elr; | |
4ed5c033 | 3852 | unsigned long next_wakeup, cur; |
bfff6873 LC |
3853 | |
3854 | BUG_ON(NULL == eli); | |
3855 | ||
bfff6873 LC |
3856 | cont_thread: |
3857 | while (true) { | |
3858 | next_wakeup = MAX_JIFFY_OFFSET; | |
3859 | ||
3860 | mutex_lock(&eli->li_list_mtx); | |
3861 | if (list_empty(&eli->li_request_list)) { | |
3862 | mutex_unlock(&eli->li_list_mtx); | |
3863 | goto exit_thread; | |
3864 | } | |
bfff6873 | 3865 | list_for_each_safe(pos, n, &eli->li_request_list) { |
e22834f0 DM |
3866 | int err = 0; |
3867 | int progress = 0; | |
bfff6873 LC |
3868 | elr = list_entry(pos, struct ext4_li_request, |
3869 | lr_request); | |
3870 | ||
e22834f0 DM |
3871 | if (time_before(jiffies, elr->lr_next_sched)) { |
3872 | if (time_before(elr->lr_next_sched, next_wakeup)) | |
3873 | next_wakeup = elr->lr_next_sched; | |
3874 | continue; | |
3875 | } | |
3876 | if (down_read_trylock(&elr->lr_super->s_umount)) { | |
3877 | if (sb_start_write_trylock(elr->lr_super)) { | |
3878 | progress = 1; | |
3879 | /* | |
3880 | * We hold sb->s_umount, sb can not | |
3881 | * be removed from the list, it is | |
3882 | * now safe to drop li_list_mtx | |
3883 | */ | |
3884 | mutex_unlock(&eli->li_list_mtx); | |
3885 | err = ext4_run_li_request(elr); | |
3886 | sb_end_write(elr->lr_super); | |
3887 | mutex_lock(&eli->li_list_mtx); | |
3888 | n = pos->next; | |
b2c78cd0 | 3889 | } |
e22834f0 DM |
3890 | up_read((&elr->lr_super->s_umount)); |
3891 | } | |
3892 | /* error, remove the lazy_init job */ | |
3893 | if (err) { | |
3894 | ext4_remove_li_request(elr); | |
3895 | continue; | |
3896 | } | |
3897 | if (!progress) { | |
3898 | elr->lr_next_sched = jiffies + | |
3899 | (prandom_u32() | |
3900 | % (EXT4_DEF_LI_MAX_START_DELAY * HZ)); | |
bfff6873 | 3901 | } |
bfff6873 LC |
3902 | if (time_before(elr->lr_next_sched, next_wakeup)) |
3903 | next_wakeup = elr->lr_next_sched; | |
3904 | } | |
3905 | mutex_unlock(&eli->li_list_mtx); | |
3906 | ||
a0acae0e | 3907 | try_to_freeze(); |
bfff6873 | 3908 | |
4ed5c033 LC |
3909 | cur = jiffies; |
3910 | if ((time_after_eq(cur, next_wakeup)) || | |
f4245bd4 | 3911 | (MAX_JIFFY_OFFSET == next_wakeup)) { |
bfff6873 LC |
3912 | cond_resched(); |
3913 | continue; | |
3914 | } | |
3915 | ||
4ed5c033 LC |
3916 | schedule_timeout_interruptible(next_wakeup - cur); |
3917 | ||
8f1f7453 ES |
3918 | if (kthread_should_stop()) { |
3919 | ext4_clear_request_list(); | |
3920 | goto exit_thread; | |
3921 | } | |
bfff6873 LC |
3922 | } |
3923 | ||
3924 | exit_thread: | |
3925 | /* | |
3926 | * It looks like the request list is empty, but we need | |
3927 | * to check it under the li_list_mtx lock, to prevent any | |
3928 | * additions into it, and of course we should lock ext4_li_mtx | |
3929 | * to atomically free the list and ext4_li_info, because at | |
3930 | * this point another ext4 filesystem could be registering | |
3931 | * new one. | |
3932 | */ | |
3933 | mutex_lock(&ext4_li_mtx); | |
3934 | mutex_lock(&eli->li_list_mtx); | |
3935 | if (!list_empty(&eli->li_request_list)) { | |
3936 | mutex_unlock(&eli->li_list_mtx); | |
3937 | mutex_unlock(&ext4_li_mtx); | |
3938 | goto cont_thread; | |
3939 | } | |
3940 | mutex_unlock(&eli->li_list_mtx); | |
bfff6873 LC |
3941 | kfree(ext4_li_info); |
3942 | ext4_li_info = NULL; | |
3943 | mutex_unlock(&ext4_li_mtx); | |
3944 | ||
3945 | return 0; | |
3946 | } | |
3947 | ||
3948 | static void ext4_clear_request_list(void) | |
3949 | { | |
3950 | struct list_head *pos, *n; | |
3951 | struct ext4_li_request *elr; | |
3952 | ||
3953 | mutex_lock(&ext4_li_info->li_list_mtx); | |
bfff6873 LC |
3954 | list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { |
3955 | elr = list_entry(pos, struct ext4_li_request, | |
3956 | lr_request); | |
3957 | ext4_remove_li_request(elr); | |
3958 | } | |
3959 | mutex_unlock(&ext4_li_info->li_list_mtx); | |
3960 | } | |
3961 | ||
3962 | static int ext4_run_lazyinit_thread(void) | |
3963 | { | |
8f1f7453 ES |
3964 | ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, |
3965 | ext4_li_info, "ext4lazyinit"); | |
3966 | if (IS_ERR(ext4_lazyinit_task)) { | |
3967 | int err = PTR_ERR(ext4_lazyinit_task); | |
bfff6873 | 3968 | ext4_clear_request_list(); |
bfff6873 LC |
3969 | kfree(ext4_li_info); |
3970 | ext4_li_info = NULL; | |
92b97816 | 3971 | printk(KERN_CRIT "EXT4-fs: error %d creating inode table " |
bfff6873 LC |
3972 | "initialization thread\n", |
3973 | err); | |
3974 | return err; | |
3975 | } | |
3976 | ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; | |
bfff6873 LC |
3977 | return 0; |
3978 | } | |
3979 | ||
3980 | /* | |
3981 | * Check whether it make sense to run itable init. thread or not. | |
3982 | * If there is at least one uninitialized inode table, return | |
3983 | * corresponding group number, else the loop goes through all | |
3984 | * groups and return total number of groups. | |
3985 | */ | |
3986 | static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) | |
3987 | { | |
3988 | ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; | |
3989 | struct ext4_group_desc *gdp = NULL; | |
3990 | ||
8844618d TT |
3991 | if (!ext4_has_group_desc_csum(sb)) |
3992 | return ngroups; | |
3993 | ||
bfff6873 LC |
3994 | for (group = 0; group < ngroups; group++) { |
3995 | gdp = ext4_get_group_desc(sb, group, NULL); | |
3996 | if (!gdp) | |
3997 | continue; | |
3998 | ||
50122847 | 3999 | if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) |
bfff6873 LC |
4000 | break; |
4001 | } | |
4002 | ||
4003 | return group; | |
4004 | } | |
4005 | ||
4006 | static int ext4_li_info_new(void) | |
4007 | { | |
4008 | struct ext4_lazy_init *eli = NULL; | |
4009 | ||
4010 | eli = kzalloc(sizeof(*eli), GFP_KERNEL); | |
4011 | if (!eli) | |
4012 | return -ENOMEM; | |
4013 | ||
bfff6873 LC |
4014 | INIT_LIST_HEAD(&eli->li_request_list); |
4015 | mutex_init(&eli->li_list_mtx); | |
4016 | ||
bfff6873 LC |
4017 | eli->li_state |= EXT4_LAZYINIT_QUIT; |
4018 | ||
4019 | ext4_li_info = eli; | |
4020 | ||
4021 | return 0; | |
4022 | } | |
4023 | ||
4024 | static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, | |
4025 | ext4_group_t start) | |
4026 | { | |
bfff6873 | 4027 | struct ext4_li_request *elr; |
bfff6873 LC |
4028 | |
4029 | elr = kzalloc(sizeof(*elr), GFP_KERNEL); | |
4030 | if (!elr) | |
4031 | return NULL; | |
4032 | ||
4033 | elr->lr_super = sb; | |
3d392b26 | 4034 | elr->lr_first_not_zeroed = start; |
21175ca4 | 4035 | if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) { |
3d392b26 TT |
4036 | elr->lr_mode = EXT4_LI_MODE_ITABLE; |
4037 | elr->lr_next_group = start; | |
21175ca4 HS |
4038 | } else { |
4039 | elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP; | |
3d392b26 | 4040 | } |
bfff6873 LC |
4041 | |
4042 | /* | |
4043 | * Randomize first schedule time of the request to | |
4044 | * spread the inode table initialization requests | |
4045 | * better. | |
4046 | */ | |
dd1f723b TT |
4047 | elr->lr_next_sched = jiffies + (prandom_u32() % |
4048 | (EXT4_DEF_LI_MAX_START_DELAY * HZ)); | |
bfff6873 LC |
4049 | return elr; |
4050 | } | |
4051 | ||
7f511862 TT |
4052 | int ext4_register_li_request(struct super_block *sb, |
4053 | ext4_group_t first_not_zeroed) | |
bfff6873 LC |
4054 | { |
4055 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
7f511862 | 4056 | struct ext4_li_request *elr = NULL; |
49598e04 | 4057 | ext4_group_t ngroups = sbi->s_groups_count; |
6c5a6cb9 | 4058 | int ret = 0; |
bfff6873 | 4059 | |
7f511862 | 4060 | mutex_lock(&ext4_li_mtx); |
51ce6511 LC |
4061 | if (sbi->s_li_request != NULL) { |
4062 | /* | |
4063 | * Reset timeout so it can be computed again, because | |
4064 | * s_li_wait_mult might have changed. | |
4065 | */ | |
4066 | sbi->s_li_request->lr_timeout = 0; | |
7f511862 | 4067 | goto out; |
51ce6511 | 4068 | } |
bfff6873 | 4069 | |
21175ca4 | 4070 | if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) && |
3d392b26 TT |
4071 | (first_not_zeroed == ngroups || sb_rdonly(sb) || |
4072 | !test_opt(sb, INIT_INODE_TABLE))) | |
7f511862 | 4073 | goto out; |
bfff6873 LC |
4074 | |
4075 | elr = ext4_li_request_new(sb, first_not_zeroed); | |
7f511862 TT |
4076 | if (!elr) { |
4077 | ret = -ENOMEM; | |
4078 | goto out; | |
4079 | } | |
bfff6873 LC |
4080 | |
4081 | if (NULL == ext4_li_info) { | |
4082 | ret = ext4_li_info_new(); | |
4083 | if (ret) | |
4084 | goto out; | |
4085 | } | |
4086 | ||
4087 | mutex_lock(&ext4_li_info->li_list_mtx); | |
4088 | list_add(&elr->lr_request, &ext4_li_info->li_request_list); | |
4089 | mutex_unlock(&ext4_li_info->li_list_mtx); | |
4090 | ||
4091 | sbi->s_li_request = elr; | |
46e4690b TM |
4092 | /* |
4093 | * set elr to NULL here since it has been inserted to | |
4094 | * the request_list and the removal and free of it is | |
4095 | * handled by ext4_clear_request_list from now on. | |
4096 | */ | |
4097 | elr = NULL; | |
bfff6873 LC |
4098 | |
4099 | if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { | |
4100 | ret = ext4_run_lazyinit_thread(); | |
4101 | if (ret) | |
4102 | goto out; | |
4103 | } | |
bfff6873 | 4104 | out: |
beed5ecb NK |
4105 | mutex_unlock(&ext4_li_mtx); |
4106 | if (ret) | |
bfff6873 | 4107 | kfree(elr); |
bfff6873 LC |
4108 | return ret; |
4109 | } | |
4110 | ||
4111 | /* | |
4112 | * We do not need to lock anything since this is called on | |
4113 | * module unload. | |
4114 | */ | |
4115 | static void ext4_destroy_lazyinit_thread(void) | |
4116 | { | |
4117 | /* | |
4118 | * If thread exited earlier | |
4119 | * there's nothing to be done. | |
4120 | */ | |
8f1f7453 | 4121 | if (!ext4_li_info || !ext4_lazyinit_task) |
bfff6873 LC |
4122 | return; |
4123 | ||
8f1f7453 | 4124 | kthread_stop(ext4_lazyinit_task); |
bfff6873 LC |
4125 | } |
4126 | ||
25ed6e8a DW |
4127 | static int set_journal_csum_feature_set(struct super_block *sb) |
4128 | { | |
4129 | int ret = 1; | |
4130 | int compat, incompat; | |
4131 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
4132 | ||
9aa5d32b | 4133 | if (ext4_has_metadata_csum(sb)) { |
db9ee220 | 4134 | /* journal checksum v3 */ |
25ed6e8a | 4135 | compat = 0; |
db9ee220 | 4136 | incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; |
25ed6e8a DW |
4137 | } else { |
4138 | /* journal checksum v1 */ | |
4139 | compat = JBD2_FEATURE_COMPAT_CHECKSUM; | |
4140 | incompat = 0; | |
4141 | } | |
4142 | ||
feb8c6d3 DW |
4143 | jbd2_journal_clear_features(sbi->s_journal, |
4144 | JBD2_FEATURE_COMPAT_CHECKSUM, 0, | |
4145 | JBD2_FEATURE_INCOMPAT_CSUM_V3 | | |
4146 | JBD2_FEATURE_INCOMPAT_CSUM_V2); | |
25ed6e8a DW |
4147 | if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { |
4148 | ret = jbd2_journal_set_features(sbi->s_journal, | |
4149 | compat, 0, | |
4150 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | | |
4151 | incompat); | |
4152 | } else if (test_opt(sb, JOURNAL_CHECKSUM)) { | |
4153 | ret = jbd2_journal_set_features(sbi->s_journal, | |
4154 | compat, 0, | |
4155 | incompat); | |
4156 | jbd2_journal_clear_features(sbi->s_journal, 0, 0, | |
4157 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); | |
4158 | } else { | |
feb8c6d3 DW |
4159 | jbd2_journal_clear_features(sbi->s_journal, 0, 0, |
4160 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); | |
25ed6e8a DW |
4161 | } |
4162 | ||
4163 | return ret; | |
4164 | } | |
4165 | ||
952fc18e TT |
4166 | /* |
4167 | * Note: calculating the overhead so we can be compatible with | |
4168 | * historical BSD practice is quite difficult in the face of | |
4169 | * clusters/bigalloc. This is because multiple metadata blocks from | |
4170 | * different block group can end up in the same allocation cluster. | |
4171 | * Calculating the exact overhead in the face of clustered allocation | |
4172 | * requires either O(all block bitmaps) in memory or O(number of block | |
4173 | * groups**2) in time. We will still calculate the superblock for | |
4174 | * older file systems --- and if we come across with a bigalloc file | |
4175 | * system with zero in s_overhead_clusters the estimate will be close to | |
4176 | * correct especially for very large cluster sizes --- but for newer | |
4177 | * file systems, it's better to calculate this figure once at mkfs | |
4178 | * time, and store it in the superblock. If the superblock value is | |
4179 | * present (even for non-bigalloc file systems), we will use it. | |
4180 | */ | |
4181 | static int count_overhead(struct super_block *sb, ext4_group_t grp, | |
4182 | char *buf) | |
4183 | { | |
4184 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
4185 | struct ext4_group_desc *gdp; | |
4186 | ext4_fsblk_t first_block, last_block, b; | |
4187 | ext4_group_t i, ngroups = ext4_get_groups_count(sb); | |
4188 | int s, j, count = 0; | |
4189 | ||
e2b911c5 | 4190 | if (!ext4_has_feature_bigalloc(sb)) |
0548bbb8 TT |
4191 | return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) + |
4192 | sbi->s_itb_per_group + 2); | |
4193 | ||
952fc18e TT |
4194 | first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + |
4195 | (grp * EXT4_BLOCKS_PER_GROUP(sb)); | |
4196 | last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; | |
4197 | for (i = 0; i < ngroups; i++) { | |
4198 | gdp = ext4_get_group_desc(sb, i, NULL); | |
4199 | b = ext4_block_bitmap(sb, gdp); | |
4200 | if (b >= first_block && b <= last_block) { | |
4201 | ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); | |
4202 | count++; | |
4203 | } | |
4204 | b = ext4_inode_bitmap(sb, gdp); | |
4205 | if (b >= first_block && b <= last_block) { | |
4206 | ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); | |
4207 | count++; | |
4208 | } | |
4209 | b = ext4_inode_table(sb, gdp); | |
4210 | if (b >= first_block && b + sbi->s_itb_per_group <= last_block) | |
4211 | for (j = 0; j < sbi->s_itb_per_group; j++, b++) { | |
4212 | int c = EXT4_B2C(sbi, b - first_block); | |
4213 | ext4_set_bit(c, buf); | |
4214 | count++; | |
4215 | } | |
4216 | if (i != grp) | |
4217 | continue; | |
4218 | s = 0; | |
4219 | if (ext4_bg_has_super(sb, grp)) { | |
4220 | ext4_set_bit(s++, buf); | |
4221 | count++; | |
4222 | } | |
c48ae41b TT |
4223 | j = ext4_bg_num_gdb(sb, grp); |
4224 | if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) { | |
4225 | ext4_error(sb, "Invalid number of block group " | |
4226 | "descriptor blocks: %d", j); | |
4227 | j = EXT4_BLOCKS_PER_GROUP(sb) - s; | |
952fc18e | 4228 | } |
c48ae41b TT |
4229 | count += j; |
4230 | for (; j > 0; j--) | |
4231 | ext4_set_bit(EXT4_B2C(sbi, s++), buf); | |
952fc18e TT |
4232 | } |
4233 | if (!count) | |
4234 | return 0; | |
4235 | return EXT4_CLUSTERS_PER_GROUP(sb) - | |
4236 | ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); | |
4237 | } | |
4238 | ||
4239 | /* | |
4240 | * Compute the overhead and stash it in sbi->s_overhead | |
4241 | */ | |
4242 | int ext4_calculate_overhead(struct super_block *sb) | |
4243 | { | |
4244 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
4245 | struct ext4_super_block *es = sbi->s_es; | |
3c816ded EW |
4246 | struct inode *j_inode; |
4247 | unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum); | |
952fc18e TT |
4248 | ext4_group_t i, ngroups = ext4_get_groups_count(sb); |
4249 | ext4_fsblk_t overhead = 0; | |
4fdb5543 | 4250 | char *buf = (char *) get_zeroed_page(GFP_NOFS); |
952fc18e | 4251 | |
952fc18e TT |
4252 | if (!buf) |
4253 | return -ENOMEM; | |
4254 | ||
4255 | /* | |
4256 | * Compute the overhead (FS structures). This is constant | |
4257 | * for a given filesystem unless the number of block groups | |
4258 | * changes so we cache the previous value until it does. | |
4259 | */ | |
4260 | ||
4261 | /* | |
4262 | * All of the blocks before first_data_block are overhead | |
4263 | */ | |
4264 | overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); | |
4265 | ||
4266 | /* | |
4267 | * Add the overhead found in each block group | |
4268 | */ | |
4269 | for (i = 0; i < ngroups; i++) { | |
4270 | int blks; | |
4271 | ||
4272 | blks = count_overhead(sb, i, buf); | |
4273 | overhead += blks; | |
4274 | if (blks) | |
4275 | memset(buf, 0, PAGE_SIZE); | |
4276 | cond_resched(); | |
4277 | } | |
3c816ded EW |
4278 | |
4279 | /* | |
4280 | * Add the internal journal blocks whether the journal has been | |
4281 | * loaded or not | |
4282 | */ | |
ee7ed3aa | 4283 | if (sbi->s_journal && !sbi->s_journal_bdev) |
ede7dc7f | 4284 | overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len); |
f1eec3b0 RH |
4285 | else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) { |
4286 | /* j_inum for internal journal is non-zero */ | |
3c816ded EW |
4287 | j_inode = ext4_get_journal_inode(sb, j_inum); |
4288 | if (j_inode) { | |
4289 | j_blocks = j_inode->i_size >> sb->s_blocksize_bits; | |
4290 | overhead += EXT4_NUM_B2C(sbi, j_blocks); | |
4291 | iput(j_inode); | |
4292 | } else { | |
4293 | ext4_msg(sb, KERN_ERR, "can't get journal size"); | |
4294 | } | |
4295 | } | |
952fc18e TT |
4296 | sbi->s_overhead = overhead; |
4297 | smp_wmb(); | |
4298 | free_page((unsigned long) buf); | |
4299 | return 0; | |
4300 | } | |
4301 | ||
b5799018 | 4302 | static void ext4_set_resv_clusters(struct super_block *sb) |
27dd4385 LC |
4303 | { |
4304 | ext4_fsblk_t resv_clusters; | |
b5799018 | 4305 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
27dd4385 | 4306 | |
30fac0f7 JK |
4307 | /* |
4308 | * There's no need to reserve anything when we aren't using extents. | |
4309 | * The space estimates are exact, there are no unwritten extents, | |
4310 | * hole punching doesn't need new metadata... This is needed especially | |
4311 | * to keep ext2/3 backward compatibility. | |
4312 | */ | |
e2b911c5 | 4313 | if (!ext4_has_feature_extents(sb)) |
b5799018 | 4314 | return; |
27dd4385 LC |
4315 | /* |
4316 | * By default we reserve 2% or 4096 clusters, whichever is smaller. | |
4317 | * This should cover the situations where we can not afford to run | |
4318 | * out of space like for example punch hole, or converting | |
556615dc | 4319 | * unwritten extents in delalloc path. In most cases such |
27dd4385 LC |
4320 | * allocation would require 1, or 2 blocks, higher numbers are |
4321 | * very rare. | |
4322 | */ | |
b5799018 TT |
4323 | resv_clusters = (ext4_blocks_count(sbi->s_es) >> |
4324 | sbi->s_cluster_bits); | |
27dd4385 LC |
4325 | |
4326 | do_div(resv_clusters, 50); | |
4327 | resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); | |
4328 | ||
b5799018 | 4329 | atomic64_set(&sbi->s_resv_clusters, resv_clusters); |
27dd4385 LC |
4330 | } |
4331 | ||
ca9b404f RA |
4332 | static const char *ext4_quota_mode(struct super_block *sb) |
4333 | { | |
4334 | #ifdef CONFIG_QUOTA | |
4335 | if (!ext4_quota_capable(sb)) | |
4336 | return "none"; | |
4337 | ||
4338 | if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb)) | |
4339 | return "journalled"; | |
4340 | else | |
4341 | return "writeback"; | |
4342 | #else | |
4343 | return "disabled"; | |
4344 | #endif | |
4345 | } | |
4346 | ||
188c299e JK |
4347 | static void ext4_setup_csum_trigger(struct super_block *sb, |
4348 | enum ext4_journal_trigger_type type, | |
4349 | void (*trigger)( | |
4350 | struct jbd2_buffer_trigger_type *type, | |
4351 | struct buffer_head *bh, | |
4352 | void *mapped_data, | |
4353 | size_t size)) | |
4354 | { | |
4355 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
4356 | ||
4357 | sbi->s_journal_triggers[type].sb = sb; | |
4358 | sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger; | |
4359 | } | |
4360 | ||
2b2d6d01 | 4361 | static int ext4_fill_super(struct super_block *sb, void *data, int silent) |
ac27a0ec | 4362 | { |
5e405595 | 4363 | struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev); |
d4c402d9 | 4364 | char *orig_data = kstrdup(data, GFP_KERNEL); |
1d0c3924 | 4365 | struct buffer_head *bh, **group_desc; |
617ba13b | 4366 | struct ext4_super_block *es = NULL; |
5aee0f8a | 4367 | struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); |
7c990728 | 4368 | struct flex_groups **flex_groups; |
617ba13b MC |
4369 | ext4_fsblk_t block; |
4370 | ext4_fsblk_t sb_block = get_sb_block(&data); | |
70bbb3e0 | 4371 | ext4_fsblk_t logical_sb_block; |
ac27a0ec | 4372 | unsigned long offset = 0; |
ac27a0ec DK |
4373 | unsigned long def_mount_opts; |
4374 | struct inode *root; | |
0390131b | 4375 | const char *descr; |
dcc7dae3 | 4376 | int ret = -ENOMEM; |
281b5995 | 4377 | int blocksize, clustersize; |
4ec11028 TT |
4378 | unsigned int db_count; |
4379 | unsigned int i; | |
ef5fd681 | 4380 | int needs_recovery, has_huge_files; |
bd81d8ee | 4381 | __u64 blocks_count; |
07aa2ea1 | 4382 | int err = 0; |
bfff6873 | 4383 | ext4_group_t first_not_zeroed; |
e6e268cb | 4384 | struct ext4_fs_context parsed_opts = {0}; |
b237e304 HS |
4385 | |
4386 | /* Set defaults for the variables that will be set during parsing */ | |
4387 | parsed_opts.journal_ioprio = DEFAULT_JOURNAL_IOPRIO; | |
4388 | parsed_opts.journal_devnum = 0; | |
196e402a | 4389 | parsed_opts.mb_optimize_scan = DEFAULT_MB_OPTIMIZE_SCAN; |
ac27a0ec | 4390 | |
5aee0f8a TT |
4391 | if ((data && !orig_data) || !sbi) |
4392 | goto out_free_base; | |
705895b6 | 4393 | |
aed9eb1b | 4394 | sbi->s_daxdev = dax_dev; |
705895b6 PE |
4395 | sbi->s_blockgroup_lock = |
4396 | kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); | |
5aee0f8a TT |
4397 | if (!sbi->s_blockgroup_lock) |
4398 | goto out_free_base; | |
4399 | ||
ac27a0ec | 4400 | sb->s_fs_info = sbi; |
2c0544b2 | 4401 | sbi->s_sb = sb; |
240799cd | 4402 | sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; |
d9c9bef1 | 4403 | sbi->s_sb_block = sb_block; |
8446fe92 CH |
4404 | sbi->s_sectors_written_start = |
4405 | part_stat_read(sb->s_bdev, sectors[STAT_WRITE]); | |
ac27a0ec | 4406 | |
9f6200bb | 4407 | /* Cleanup superblock name */ |
ec3904dc | 4408 | strreplace(sb->s_id, '/', '!'); |
9f6200bb | 4409 | |
07aa2ea1 | 4410 | /* -EINVAL is default */ |
dcc7dae3 | 4411 | ret = -EINVAL; |
617ba13b | 4412 | blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); |
ac27a0ec | 4413 | if (!blocksize) { |
b31e1552 | 4414 | ext4_msg(sb, KERN_ERR, "unable to set blocksize"); |
ac27a0ec DK |
4415 | goto out_fail; |
4416 | } | |
4417 | ||
4418 | /* | |
617ba13b | 4419 | * The ext4 superblock will not be buffer aligned for other than 1kB |
ac27a0ec DK |
4420 | * block sizes. We need to calculate the offset from buffer start. |
4421 | */ | |
617ba13b | 4422 | if (blocksize != EXT4_MIN_BLOCK_SIZE) { |
70bbb3e0 AM |
4423 | logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; |
4424 | offset = do_div(logical_sb_block, blocksize); | |
ac27a0ec | 4425 | } else { |
70bbb3e0 | 4426 | logical_sb_block = sb_block; |
ac27a0ec DK |
4427 | } |
4428 | ||
8394a6ab | 4429 | bh = ext4_sb_bread_unmovable(sb, logical_sb_block); |
4430 | if (IS_ERR(bh)) { | |
b31e1552 | 4431 | ext4_msg(sb, KERN_ERR, "unable to read superblock"); |
8394a6ab | 4432 | ret = PTR_ERR(bh); |
ac27a0ec DK |
4433 | goto out_fail; |
4434 | } | |
4435 | /* | |
4436 | * Note: s_es must be initialized as soon as possible because | |
617ba13b | 4437 | * some ext4 macro-instructions depend on its value |
ac27a0ec | 4438 | */ |
2716b802 | 4439 | es = (struct ext4_super_block *) (bh->b_data + offset); |
ac27a0ec DK |
4440 | sbi->s_es = es; |
4441 | sb->s_magic = le16_to_cpu(es->s_magic); | |
617ba13b MC |
4442 | if (sb->s_magic != EXT4_SUPER_MAGIC) |
4443 | goto cantfind_ext4; | |
afc32f7e | 4444 | sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); |
ac27a0ec | 4445 | |
feb0ab32 | 4446 | /* Warn if metadata_csum and gdt_csum are both set. */ |
e2b911c5 DW |
4447 | if (ext4_has_feature_metadata_csum(sb) && |
4448 | ext4_has_feature_gdt_csum(sb)) | |
363307e6 | 4449 | ext4_warning(sb, "metadata_csum and uninit_bg are " |
feb0ab32 DW |
4450 | "redundant flags; please run fsck."); |
4451 | ||
d25425f8 DW |
4452 | /* Check for a known checksum algorithm */ |
4453 | if (!ext4_verify_csum_type(sb, es)) { | |
4454 | ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " | |
4455 | "unknown checksum algorithm."); | |
4456 | silent = 1; | |
4457 | goto cantfind_ext4; | |
4458 | } | |
02f310fc JK |
4459 | ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE, |
4460 | ext4_orphan_file_block_trigger); | |
d25425f8 | 4461 | |
0441984a | 4462 | /* Load the checksum driver */ |
a45403b5 TT |
4463 | sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); |
4464 | if (IS_ERR(sbi->s_chksum_driver)) { | |
4465 | ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); | |
4466 | ret = PTR_ERR(sbi->s_chksum_driver); | |
4467 | sbi->s_chksum_driver = NULL; | |
4468 | goto failed_mount; | |
0441984a DW |
4469 | } |
4470 | ||
a9c47317 DW |
4471 | /* Check superblock checksum */ |
4472 | if (!ext4_superblock_csum_verify(sb, es)) { | |
4473 | ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " | |
4474 | "invalid superblock checksum. Run e2fsck?"); | |
4475 | silent = 1; | |
6a797d27 | 4476 | ret = -EFSBADCRC; |
a9c47317 DW |
4477 | goto cantfind_ext4; |
4478 | } | |
4479 | ||
4480 | /* Precompute checksum seed for all metadata */ | |
e2b911c5 | 4481 | if (ext4_has_feature_csum_seed(sb)) |
8c81bd8f | 4482 | sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); |
dec214d0 | 4483 | else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb)) |
a9c47317 DW |
4484 | sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, |
4485 | sizeof(es->s_uuid)); | |
4486 | ||
ac27a0ec DK |
4487 | /* Set defaults before we parse the mount options */ |
4488 | def_mount_opts = le32_to_cpu(es->s_default_mount_opts); | |
fd8c37ec | 4489 | set_opt(sb, INIT_INODE_TABLE); |
617ba13b | 4490 | if (def_mount_opts & EXT4_DEFM_DEBUG) |
fd8c37ec | 4491 | set_opt(sb, DEBUG); |
87f26807 | 4492 | if (def_mount_opts & EXT4_DEFM_BSDGROUPS) |
fd8c37ec | 4493 | set_opt(sb, GRPID); |
617ba13b | 4494 | if (def_mount_opts & EXT4_DEFM_UID16) |
fd8c37ec | 4495 | set_opt(sb, NO_UID32); |
ea663336 | 4496 | /* xattr user namespace & acls are now defaulted on */ |
ea663336 | 4497 | set_opt(sb, XATTR_USER); |
03010a33 | 4498 | #ifdef CONFIG_EXT4_FS_POSIX_ACL |
ea663336 | 4499 | set_opt(sb, POSIX_ACL); |
2e7842b8 | 4500 | #endif |
995a3ed6 HS |
4501 | if (ext4_has_feature_fast_commit(sb)) |
4502 | set_opt2(sb, JOURNAL_FAST_COMMIT); | |
98c1a759 DW |
4503 | /* don't forget to enable journal_csum when metadata_csum is enabled. */ |
4504 | if (ext4_has_metadata_csum(sb)) | |
4505 | set_opt(sb, JOURNAL_CHECKSUM); | |
4506 | ||
617ba13b | 4507 | if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) |
fd8c37ec | 4508 | set_opt(sb, JOURNAL_DATA); |
617ba13b | 4509 | else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) |
fd8c37ec | 4510 | set_opt(sb, ORDERED_DATA); |
617ba13b | 4511 | else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) |
fd8c37ec | 4512 | set_opt(sb, WRITEBACK_DATA); |
617ba13b MC |
4513 | |
4514 | if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) | |
fd8c37ec | 4515 | set_opt(sb, ERRORS_PANIC); |
bb4f397a | 4516 | else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) |
fd8c37ec | 4517 | set_opt(sb, ERRORS_CONT); |
bb4f397a | 4518 | else |
fd8c37ec | 4519 | set_opt(sb, ERRORS_RO); |
45f1a9c3 DW |
4520 | /* block_validity enabled by default; disable with noblock_validity */ |
4521 | set_opt(sb, BLOCK_VALIDITY); | |
8b67f04a | 4522 | if (def_mount_opts & EXT4_DEFM_DISCARD) |
fd8c37ec | 4523 | set_opt(sb, DISCARD); |
ac27a0ec | 4524 | |
08cefc7a EB |
4525 | sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); |
4526 | sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); | |
30773840 TT |
4527 | sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; |
4528 | sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; | |
4529 | sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; | |
ac27a0ec | 4530 | |
8b67f04a | 4531 | if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) |
fd8c37ec | 4532 | set_opt(sb, BARRIER); |
ac27a0ec | 4533 | |
dd919b98 AK |
4534 | /* |
4535 | * enable delayed allocation by default | |
4536 | * Use -o nodelalloc to turn it off | |
4537 | */ | |
bc0b75f7 | 4538 | if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && |
8b67f04a | 4539 | ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) |
fd8c37ec | 4540 | set_opt(sb, DELALLOC); |
dd919b98 | 4541 | |
51ce6511 LC |
4542 | /* |
4543 | * set default s_li_wait_mult for lazyinit, for the case there is | |
4544 | * no mount option specified. | |
4545 | */ | |
4546 | sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; | |
4547 | ||
c9200760 TT |
4548 | if (le32_to_cpu(es->s_log_block_size) > |
4549 | (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { | |
4f97a681 | 4550 | ext4_msg(sb, KERN_ERR, |
c9200760 TT |
4551 | "Invalid log block size: %u", |
4552 | le32_to_cpu(es->s_log_block_size)); | |
4f97a681 TT |
4553 | goto failed_mount; |
4554 | } | |
c9200760 TT |
4555 | if (le32_to_cpu(es->s_log_cluster_size) > |
4556 | (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { | |
4f97a681 | 4557 | ext4_msg(sb, KERN_ERR, |
c9200760 TT |
4558 | "Invalid log cluster size: %u", |
4559 | le32_to_cpu(es->s_log_cluster_size)); | |
4f97a681 TT |
4560 | goto failed_mount; |
4561 | } | |
4562 | ||
c9200760 TT |
4563 | blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); |
4564 | ||
4565 | if (blocksize == PAGE_SIZE) | |
4566 | set_opt(sb, DIOREAD_NOLOCK); | |
4f97a681 | 4567 | |
9803387c TT |
4568 | if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { |
4569 | sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; | |
4570 | sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; | |
4571 | } else { | |
4572 | sbi->s_inode_size = le16_to_cpu(es->s_inode_size); | |
4573 | sbi->s_first_ino = le32_to_cpu(es->s_first_ino); | |
4574 | if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) { | |
4575 | ext4_msg(sb, KERN_ERR, "invalid first ino: %u", | |
4576 | sbi->s_first_ino); | |
4577 | goto failed_mount; | |
4578 | } | |
4579 | if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || | |
4580 | (!is_power_of_2(sbi->s_inode_size)) || | |
4581 | (sbi->s_inode_size > blocksize)) { | |
4582 | ext4_msg(sb, KERN_ERR, | |
4583 | "unsupported inode size: %d", | |
4584 | sbi->s_inode_size); | |
4f97a681 | 4585 | ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize); |
9803387c TT |
4586 | goto failed_mount; |
4587 | } | |
4588 | /* | |
4589 | * i_atime_extra is the last extra field available for | |
4590 | * [acm]times in struct ext4_inode. Checking for that | |
4591 | * field should suffice to ensure we have extra space | |
4592 | * for all three. | |
4593 | */ | |
4594 | if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) + | |
4595 | sizeof(((struct ext4_inode *)0)->i_atime_extra)) { | |
4596 | sb->s_time_gran = 1; | |
4597 | sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX; | |
4598 | } else { | |
4599 | sb->s_time_gran = NSEC_PER_SEC; | |
4600 | sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX; | |
4601 | } | |
4602 | sb->s_time_min = EXT4_TIMESTAMP_MIN; | |
4603 | } | |
4604 | if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { | |
4605 | sbi->s_want_extra_isize = sizeof(struct ext4_inode) - | |
4606 | EXT4_GOOD_OLD_INODE_SIZE; | |
4607 | if (ext4_has_feature_extra_isize(sb)) { | |
4608 | unsigned v, max = (sbi->s_inode_size - | |
4609 | EXT4_GOOD_OLD_INODE_SIZE); | |
4610 | ||
4611 | v = le16_to_cpu(es->s_want_extra_isize); | |
4612 | if (v > max) { | |
4613 | ext4_msg(sb, KERN_ERR, | |
4614 | "bad s_want_extra_isize: %d", v); | |
4615 | goto failed_mount; | |
4616 | } | |
4617 | if (sbi->s_want_extra_isize < v) | |
4618 | sbi->s_want_extra_isize = v; | |
4619 | ||
4620 | v = le16_to_cpu(es->s_min_extra_isize); | |
4621 | if (v > max) { | |
4622 | ext4_msg(sb, KERN_ERR, | |
4623 | "bad s_min_extra_isize: %d", v); | |
4624 | goto failed_mount; | |
4625 | } | |
4626 | if (sbi->s_want_extra_isize < v) | |
4627 | sbi->s_want_extra_isize = v; | |
4628 | } | |
4629 | } | |
4630 | ||
5aee0f8a TT |
4631 | if (sbi->s_es->s_mount_opts[0]) { |
4632 | char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts, | |
4633 | sizeof(sbi->s_es->s_mount_opts), | |
4634 | GFP_KERNEL); | |
4635 | if (!s_mount_opts) | |
4636 | goto failed_mount; | |
b237e304 | 4637 | if (!parse_options(s_mount_opts, sb, &parsed_opts, 0)) { |
5aee0f8a TT |
4638 | ext4_msg(sb, KERN_WARNING, |
4639 | "failed to parse options in superblock: %s", | |
4640 | s_mount_opts); | |
4641 | } | |
4642 | kfree(s_mount_opts); | |
8b67f04a | 4643 | } |
5a916be1 | 4644 | sbi->s_def_mount_opt = sbi->s_mount_opt; |
b237e304 | 4645 | if (!parse_options((char *) data, sb, &parsed_opts, 0)) |
ac27a0ec DK |
4646 | goto failed_mount; |
4647 | ||
c83ad55e | 4648 | #ifdef CONFIG_UNICODE |
f8f4acb6 | 4649 | if (ext4_has_feature_casefold(sb) && !sb->s_encoding) { |
c83ad55e GKB |
4650 | const struct ext4_sb_encodings *encoding_info; |
4651 | struct unicode_map *encoding; | |
4652 | __u16 encoding_flags; | |
4653 | ||
c83ad55e GKB |
4654 | if (ext4_sb_read_encoding(es, &encoding_info, |
4655 | &encoding_flags)) { | |
4656 | ext4_msg(sb, KERN_ERR, | |
4657 | "Encoding requested by superblock is unknown"); | |
4658 | goto failed_mount; | |
4659 | } | |
4660 | ||
4661 | encoding = utf8_load(encoding_info->version); | |
4662 | if (IS_ERR(encoding)) { | |
4663 | ext4_msg(sb, KERN_ERR, | |
4664 | "can't mount with superblock charset: %s-%s " | |
4665 | "not supported by the kernel. flags: 0x%x.", | |
4666 | encoding_info->name, encoding_info->version, | |
4667 | encoding_flags); | |
4668 | goto failed_mount; | |
4669 | } | |
4670 | ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: " | |
4671 | "%s-%s with flags 0x%hx", encoding_info->name, | |
4672 | encoding_info->version?:"\b", encoding_flags); | |
4673 | ||
f8f4acb6 DR |
4674 | sb->s_encoding = encoding; |
4675 | sb->s_encoding_flags = encoding_flags; | |
c83ad55e GKB |
4676 | } |
4677 | #endif | |
4678 | ||
56889787 | 4679 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { |
556e0319 | 4680 | printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n"); |
781c036b | 4681 | /* can't mount with both data=journal and dioread_nolock. */ |
244adf64 | 4682 | clear_opt(sb, DIOREAD_NOLOCK); |
556e0319 | 4683 | clear_opt2(sb, JOURNAL_FAST_COMMIT); |
56889787 TT |
4684 | if (test_opt2(sb, EXPLICIT_DELALLOC)) { |
4685 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
4686 | "both data=journal and delalloc"); | |
4687 | goto failed_mount; | |
4688 | } | |
fc626fe3 | 4689 | if (test_opt(sb, DAX_ALWAYS)) { |
923ae0ff RZ |
4690 | ext4_msg(sb, KERN_ERR, "can't mount with " |
4691 | "both data=journal and dax"); | |
4692 | goto failed_mount; | |
4693 | } | |
73b92a2a SK |
4694 | if (ext4_has_feature_encrypt(sb)) { |
4695 | ext4_msg(sb, KERN_WARNING, | |
4696 | "encrypted files will use data=ordered " | |
4697 | "instead of data journaling mode"); | |
4698 | } | |
56889787 TT |
4699 | if (test_opt(sb, DELALLOC)) |
4700 | clear_opt(sb, DELALLOC); | |
001e4a87 TH |
4701 | } else { |
4702 | sb->s_iflags |= SB_I_CGROUPWB; | |
56889787 TT |
4703 | } |
4704 | ||
1751e8a6 LT |
4705 | sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | |
4706 | (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); | |
ac27a0ec | 4707 | |
617ba13b | 4708 | if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && |
e2b911c5 DW |
4709 | (ext4_has_compat_features(sb) || |
4710 | ext4_has_ro_compat_features(sb) || | |
4711 | ext4_has_incompat_features(sb))) | |
b31e1552 ES |
4712 | ext4_msg(sb, KERN_WARNING, |
4713 | "feature flags set on rev 0 fs, " | |
4714 | "running e2fsck is recommended"); | |
469108ff | 4715 | |
ed3654eb TT |
4716 | if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { |
4717 | set_opt2(sb, HURD_COMPAT); | |
e2b911c5 | 4718 | if (ext4_has_feature_64bit(sb)) { |
ed3654eb TT |
4719 | ext4_msg(sb, KERN_ERR, |
4720 | "The Hurd can't support 64-bit file systems"); | |
4721 | goto failed_mount; | |
4722 | } | |
dec214d0 TE |
4723 | |
4724 | /* | |
4725 | * ea_inode feature uses l_i_version field which is not | |
4726 | * available in HURD_COMPAT mode. | |
4727 | */ | |
4728 | if (ext4_has_feature_ea_inode(sb)) { | |
4729 | ext4_msg(sb, KERN_ERR, | |
4730 | "ea_inode feature is not supported for Hurd"); | |
4731 | goto failed_mount; | |
4732 | } | |
ed3654eb TT |
4733 | } |
4734 | ||
2035e776 TT |
4735 | if (IS_EXT2_SB(sb)) { |
4736 | if (ext2_feature_set_ok(sb)) | |
4737 | ext4_msg(sb, KERN_INFO, "mounting ext2 file system " | |
4738 | "using the ext4 subsystem"); | |
4739 | else { | |
0d9366d6 ES |
4740 | /* |
4741 | * If we're probing be silent, if this looks like | |
4742 | * it's actually an ext[34] filesystem. | |
4743 | */ | |
4744 | if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) | |
4745 | goto failed_mount; | |
2035e776 TT |
4746 | ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " |
4747 | "to feature incompatibilities"); | |
4748 | goto failed_mount; | |
4749 | } | |
4750 | } | |
4751 | ||
4752 | if (IS_EXT3_SB(sb)) { | |
4753 | if (ext3_feature_set_ok(sb)) | |
4754 | ext4_msg(sb, KERN_INFO, "mounting ext3 file system " | |
4755 | "using the ext4 subsystem"); | |
4756 | else { | |
0d9366d6 ES |
4757 | /* |
4758 | * If we're probing be silent, if this looks like | |
4759 | * it's actually an ext4 filesystem. | |
4760 | */ | |
4761 | if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) | |
4762 | goto failed_mount; | |
2035e776 TT |
4763 | ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " |
4764 | "to feature incompatibilities"); | |
4765 | goto failed_mount; | |
4766 | } | |
4767 | } | |
4768 | ||
ac27a0ec DK |
4769 | /* |
4770 | * Check feature flags regardless of the revision level, since we | |
4771 | * previously didn't change the revision level when setting the flags, | |
4772 | * so there is a chance incompat flags are set on a rev 0 filesystem. | |
4773 | */ | |
bc98a42c | 4774 | if (!ext4_feature_set_ok(sb, (sb_rdonly(sb)))) |
ac27a0ec | 4775 | goto failed_mount; |
a13fb1a4 | 4776 | |
5b9554dc TT |
4777 | if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) { |
4778 | ext4_msg(sb, KERN_ERR, | |
4779 | "Number of reserved GDT blocks insanely large: %d", | |
4780 | le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks)); | |
4781 | goto failed_mount; | |
4782 | } | |
4783 | ||
bdd3c50d CH |
4784 | if (dax_supported(dax_dev, sb->s_bdev, blocksize, 0, |
4785 | bdev_nr_sectors(sb->s_bdev))) | |
a8ab6d38 IW |
4786 | set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags); |
4787 | ||
fc626fe3 | 4788 | if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) { |
559db4c6 RZ |
4789 | if (ext4_has_feature_inline_data(sb)) { |
4790 | ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem" | |
4791 | " that may contain inline data"); | |
361d24d4 | 4792 | goto failed_mount; |
559db4c6 | 4793 | } |
a8ab6d38 | 4794 | if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) { |
24f3478d | 4795 | ext4_msg(sb, KERN_ERR, |
361d24d4 ES |
4796 | "DAX unsupported by block device."); |
4797 | goto failed_mount; | |
24f3478d | 4798 | } |
923ae0ff RZ |
4799 | } |
4800 | ||
e2b911c5 | 4801 | if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { |
6ddb2447 TT |
4802 | ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", |
4803 | es->s_encryption_level); | |
4804 | goto failed_mount; | |
4805 | } | |
4806 | ||
ac27a0ec | 4807 | if (sb->s_blocksize != blocksize) { |
afd09b61 AM |
4808 | /* |
4809 | * bh must be released before kill_bdev(), otherwise | |
4810 | * it won't be freed and its page also. kill_bdev() | |
4811 | * is called by sb_set_blocksize(). | |
4812 | */ | |
4813 | brelse(bh); | |
ce40733c AK |
4814 | /* Validate the filesystem blocksize */ |
4815 | if (!sb_set_blocksize(sb, blocksize)) { | |
b31e1552 | 4816 | ext4_msg(sb, KERN_ERR, "bad block size %d", |
ce40733c | 4817 | blocksize); |
afd09b61 | 4818 | bh = NULL; |
ac27a0ec DK |
4819 | goto failed_mount; |
4820 | } | |
4821 | ||
70bbb3e0 AM |
4822 | logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; |
4823 | offset = do_div(logical_sb_block, blocksize); | |
8394a6ab | 4824 | bh = ext4_sb_bread_unmovable(sb, logical_sb_block); |
4825 | if (IS_ERR(bh)) { | |
b31e1552 ES |
4826 | ext4_msg(sb, KERN_ERR, |
4827 | "Can't read superblock on 2nd try"); | |
8394a6ab | 4828 | ret = PTR_ERR(bh); |
4829 | bh = NULL; | |
ac27a0ec DK |
4830 | goto failed_mount; |
4831 | } | |
2716b802 | 4832 | es = (struct ext4_super_block *)(bh->b_data + offset); |
ac27a0ec | 4833 | sbi->s_es = es; |
617ba13b | 4834 | if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { |
b31e1552 ES |
4835 | ext4_msg(sb, KERN_ERR, |
4836 | "Magic mismatch, very weird!"); | |
ac27a0ec DK |
4837 | goto failed_mount; |
4838 | } | |
4839 | } | |
4840 | ||
e2b911c5 | 4841 | has_huge_files = ext4_has_feature_huge_file(sb); |
f287a1a5 TT |
4842 | sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, |
4843 | has_huge_files); | |
4844 | sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); | |
ac27a0ec | 4845 | |
0d1ee42f | 4846 | sbi->s_desc_size = le16_to_cpu(es->s_desc_size); |
e2b911c5 | 4847 | if (ext4_has_feature_64bit(sb)) { |
8fadc143 | 4848 | if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || |
0d1ee42f | 4849 | sbi->s_desc_size > EXT4_MAX_DESC_SIZE || |
d8ea6cf8 | 4850 | !is_power_of_2(sbi->s_desc_size)) { |
b31e1552 ES |
4851 | ext4_msg(sb, KERN_ERR, |
4852 | "unsupported descriptor size %lu", | |
0d1ee42f AR |
4853 | sbi->s_desc_size); |
4854 | goto failed_mount; | |
4855 | } | |
4856 | } else | |
4857 | sbi->s_desc_size = EXT4_MIN_DESC_SIZE; | |
0b8e58a1 | 4858 | |
ac27a0ec | 4859 | sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); |
ac27a0ec | 4860 | sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); |
0b8e58a1 | 4861 | |
617ba13b | 4862 | sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); |
ac27a0ec | 4863 | if (sbi->s_inodes_per_block == 0) |
617ba13b | 4864 | goto cantfind_ext4; |
cd6bb35b TT |
4865 | if (sbi->s_inodes_per_group < sbi->s_inodes_per_block || |
4866 | sbi->s_inodes_per_group > blocksize * 8) { | |
4867 | ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n", | |
b9c538da | 4868 | sbi->s_inodes_per_group); |
cd6bb35b TT |
4869 | goto failed_mount; |
4870 | } | |
ac27a0ec DK |
4871 | sbi->s_itb_per_group = sbi->s_inodes_per_group / |
4872 | sbi->s_inodes_per_block; | |
0d1ee42f | 4873 | sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); |
ac27a0ec DK |
4874 | sbi->s_sbh = bh; |
4875 | sbi->s_mount_state = le16_to_cpu(es->s_state); | |
e57aa839 FW |
4876 | sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); |
4877 | sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); | |
0b8e58a1 | 4878 | |
2b2d6d01 | 4879 | for (i = 0; i < 4; i++) |
ac27a0ec DK |
4880 | sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); |
4881 | sbi->s_def_hash_version = es->s_def_hash_version; | |
e2b911c5 | 4882 | if (ext4_has_feature_dir_index(sb)) { |
23301410 TT |
4883 | i = le32_to_cpu(es->s_flags); |
4884 | if (i & EXT2_FLAGS_UNSIGNED_HASH) | |
4885 | sbi->s_hash_unsigned = 3; | |
4886 | else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { | |
f99b2589 | 4887 | #ifdef __CHAR_UNSIGNED__ |
bc98a42c | 4888 | if (!sb_rdonly(sb)) |
23301410 TT |
4889 | es->s_flags |= |
4890 | cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); | |
4891 | sbi->s_hash_unsigned = 3; | |
f99b2589 | 4892 | #else |
bc98a42c | 4893 | if (!sb_rdonly(sb)) |
23301410 TT |
4894 | es->s_flags |= |
4895 | cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); | |
f99b2589 | 4896 | #endif |
23301410 | 4897 | } |
f99b2589 | 4898 | } |
ac27a0ec | 4899 | |
281b5995 TT |
4900 | /* Handle clustersize */ |
4901 | clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); | |
ef5fd681 | 4902 | if (ext4_has_feature_bigalloc(sb)) { |
281b5995 TT |
4903 | if (clustersize < blocksize) { |
4904 | ext4_msg(sb, KERN_ERR, | |
4905 | "cluster size (%d) smaller than " | |
4906 | "block size (%d)", clustersize, blocksize); | |
4907 | goto failed_mount; | |
4908 | } | |
4909 | sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - | |
4910 | le32_to_cpu(es->s_log_block_size); | |
4911 | sbi->s_clusters_per_group = | |
4912 | le32_to_cpu(es->s_clusters_per_group); | |
4913 | if (sbi->s_clusters_per_group > blocksize * 8) { | |
4914 | ext4_msg(sb, KERN_ERR, | |
4915 | "#clusters per group too big: %lu", | |
4916 | sbi->s_clusters_per_group); | |
4917 | goto failed_mount; | |
4918 | } | |
4919 | if (sbi->s_blocks_per_group != | |
4920 | (sbi->s_clusters_per_group * (clustersize / blocksize))) { | |
4921 | ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " | |
4922 | "clusters per group (%lu) inconsistent", | |
4923 | sbi->s_blocks_per_group, | |
4924 | sbi->s_clusters_per_group); | |
4925 | goto failed_mount; | |
4926 | } | |
4927 | } else { | |
4928 | if (clustersize != blocksize) { | |
bfe0a5f4 TT |
4929 | ext4_msg(sb, KERN_ERR, |
4930 | "fragment/cluster size (%d) != " | |
4931 | "block size (%d)", clustersize, blocksize); | |
4932 | goto failed_mount; | |
281b5995 TT |
4933 | } |
4934 | if (sbi->s_blocks_per_group > blocksize * 8) { | |
4935 | ext4_msg(sb, KERN_ERR, | |
4936 | "#blocks per group too big: %lu", | |
4937 | sbi->s_blocks_per_group); | |
4938 | goto failed_mount; | |
4939 | } | |
4940 | sbi->s_clusters_per_group = sbi->s_blocks_per_group; | |
4941 | sbi->s_cluster_bits = 0; | |
ac27a0ec | 4942 | } |
281b5995 TT |
4943 | sbi->s_cluster_ratio = clustersize / blocksize; |
4944 | ||
960fd856 TT |
4945 | /* Do we have standard group size of clustersize * 8 blocks ? */ |
4946 | if (sbi->s_blocks_per_group == clustersize << 3) | |
4947 | set_opt2(sb, STD_GROUP_SIZE); | |
4948 | ||
bf43d84b ES |
4949 | /* |
4950 | * Test whether we have more sectors than will fit in sector_t, | |
4951 | * and whether the max offset is addressable by the page cache. | |
4952 | */ | |
5a9ae68a | 4953 | err = generic_check_addressable(sb->s_blocksize_bits, |
30ca22c7 | 4954 | ext4_blocks_count(es)); |
5a9ae68a | 4955 | if (err) { |
b31e1552 | 4956 | ext4_msg(sb, KERN_ERR, "filesystem" |
bf43d84b | 4957 | " too large to mount safely on this system"); |
ac27a0ec DK |
4958 | goto failed_mount; |
4959 | } | |
4960 | ||
617ba13b MC |
4961 | if (EXT4_BLOCKS_PER_GROUP(sb) == 0) |
4962 | goto cantfind_ext4; | |
e7c95593 | 4963 | |
0f2ddca6 | 4964 | /* check blocks count against device size */ |
5513b241 | 4965 | blocks_count = sb_bdev_nr_blocks(sb); |
0f2ddca6 | 4966 | if (blocks_count && ext4_blocks_count(es) > blocks_count) { |
b31e1552 ES |
4967 | ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " |
4968 | "exceeds size of device (%llu blocks)", | |
0f2ddca6 FTN |
4969 | ext4_blocks_count(es), blocks_count); |
4970 | goto failed_mount; | |
4971 | } | |
4972 | ||
0b8e58a1 AD |
4973 | /* |
4974 | * It makes no sense for the first data block to be beyond the end | |
4975 | * of the filesystem. | |
4976 | */ | |
4977 | if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { | |
5635a62b | 4978 | ext4_msg(sb, KERN_WARNING, "bad geometry: first data " |
b31e1552 ES |
4979 | "block %u is beyond end of filesystem (%llu)", |
4980 | le32_to_cpu(es->s_first_data_block), | |
4981 | ext4_blocks_count(es)); | |
e7c95593 ES |
4982 | goto failed_mount; |
4983 | } | |
bfe0a5f4 TT |
4984 | if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) && |
4985 | (sbi->s_cluster_ratio == 1)) { | |
4986 | ext4_msg(sb, KERN_WARNING, "bad geometry: first data " | |
4987 | "block is 0 with a 1k block and cluster size"); | |
4988 | goto failed_mount; | |
4989 | } | |
4990 | ||
bd81d8ee LV |
4991 | blocks_count = (ext4_blocks_count(es) - |
4992 | le32_to_cpu(es->s_first_data_block) + | |
4993 | EXT4_BLOCKS_PER_GROUP(sb) - 1); | |
4994 | do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); | |
4ec11028 | 4995 | if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { |
df41460a | 4996 | ext4_msg(sb, KERN_WARNING, "groups count too large: %llu " |
4ec11028 | 4997 | "(block count %llu, first data block %u, " |
df41460a | 4998 | "blocks per group %lu)", blocks_count, |
4ec11028 TT |
4999 | ext4_blocks_count(es), |
5000 | le32_to_cpu(es->s_first_data_block), | |
5001 | EXT4_BLOCKS_PER_GROUP(sb)); | |
5002 | goto failed_mount; | |
5003 | } | |
bd81d8ee | 5004 | sbi->s_groups_count = blocks_count; |
fb0a387d ES |
5005 | sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, |
5006 | (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); | |
9e463084 TT |
5007 | if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) != |
5008 | le32_to_cpu(es->s_inodes_count)) { | |
5009 | ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu", | |
5010 | le32_to_cpu(es->s_inodes_count), | |
5011 | ((u64)sbi->s_groups_count * sbi->s_inodes_per_group)); | |
5012 | ret = -EINVAL; | |
5013 | goto failed_mount; | |
5014 | } | |
617ba13b MC |
5015 | db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / |
5016 | EXT4_DESC_PER_BLOCK(sb); | |
3a4b77cd | 5017 | if (ext4_has_feature_meta_bg(sb)) { |
2ba3e6e8 | 5018 | if (le32_to_cpu(es->s_first_meta_bg) > db_count) { |
3a4b77cd EG |
5019 | ext4_msg(sb, KERN_WARNING, |
5020 | "first meta block group too large: %u " | |
5021 | "(group descriptor block count %u)", | |
5022 | le32_to_cpu(es->s_first_meta_bg), db_count); | |
5023 | goto failed_mount; | |
5024 | } | |
5025 | } | |
1d0c3924 TT |
5026 | rcu_assign_pointer(sbi->s_group_desc, |
5027 | kvmalloc_array(db_count, | |
5028 | sizeof(struct buffer_head *), | |
5029 | GFP_KERNEL)); | |
ac27a0ec | 5030 | if (sbi->s_group_desc == NULL) { |
b31e1552 | 5031 | ext4_msg(sb, KERN_ERR, "not enough memory"); |
2cde417d | 5032 | ret = -ENOMEM; |
ac27a0ec DK |
5033 | goto failed_mount; |
5034 | } | |
5035 | ||
705895b6 | 5036 | bgl_lock_init(sbi->s_blockgroup_lock); |
ac27a0ec | 5037 | |
85c8f176 AP |
5038 | /* Pre-read the descriptors into the buffer cache */ |
5039 | for (i = 0; i < db_count; i++) { | |
5040 | block = descriptor_loc(sb, logical_sb_block, i); | |
5df1d412 | 5041 | ext4_sb_breadahead_unmovable(sb, block); |
85c8f176 AP |
5042 | } |
5043 | ||
ac27a0ec | 5044 | for (i = 0; i < db_count; i++) { |
1d0c3924 TT |
5045 | struct buffer_head *bh; |
5046 | ||
70bbb3e0 | 5047 | block = descriptor_loc(sb, logical_sb_block, i); |
8394a6ab | 5048 | bh = ext4_sb_bread_unmovable(sb, block); |
5049 | if (IS_ERR(bh)) { | |
b31e1552 ES |
5050 | ext4_msg(sb, KERN_ERR, |
5051 | "can't read group descriptor %d", i); | |
ac27a0ec | 5052 | db_count = i; |
8394a6ab | 5053 | ret = PTR_ERR(bh); |
ac27a0ec DK |
5054 | goto failed_mount2; |
5055 | } | |
1d0c3924 TT |
5056 | rcu_read_lock(); |
5057 | rcu_dereference(sbi->s_group_desc)[i] = bh; | |
5058 | rcu_read_unlock(); | |
ac27a0ec | 5059 | } |
44de022c | 5060 | sbi->s_gdb_count = db_count; |
829fa70d | 5061 | if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) { |
b31e1552 | 5062 | ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); |
6a797d27 | 5063 | ret = -EFSCORRUPTED; |
f9ae9cf5 | 5064 | goto failed_mount2; |
ac27a0ec | 5065 | } |
772cb7c8 | 5066 | |
235699a8 | 5067 | timer_setup(&sbi->s_err_report, print_daily_error_info, 0); |
c92dc856 JK |
5068 | spin_lock_init(&sbi->s_error_lock); |
5069 | INIT_WORK(&sbi->s_error_work, flush_stashed_error_work); | |
04496411 | 5070 | |
a75ae78f | 5071 | /* Register extent status tree shrinker */ |
eb68d0e2 | 5072 | if (ext4_es_register_shrinker(sbi)) |
ce7e010a | 5073 | goto failed_mount3; |
ce7e010a | 5074 | |
c9de560d | 5075 | sbi->s_stripe = ext4_get_stripe_size(sbi); |
67a5da56 | 5076 | sbi->s_extent_max_zeroout_kb = 32; |
c9de560d | 5077 | |
f9ae9cf5 TT |
5078 | /* |
5079 | * set up enough so that it can read an inode | |
5080 | */ | |
f6e63f90 | 5081 | sb->s_op = &ext4_sops; |
617ba13b MC |
5082 | sb->s_export_op = &ext4_export_ops; |
5083 | sb->s_xattr = ext4_xattr_handlers; | |
643fa961 | 5084 | #ifdef CONFIG_FS_ENCRYPTION |
a7550b30 | 5085 | sb->s_cop = &ext4_cryptops; |
ffcc4182 | 5086 | #endif |
c93d8f88 EB |
5087 | #ifdef CONFIG_FS_VERITY |
5088 | sb->s_vop = &ext4_verityops; | |
5089 | #endif | |
ac27a0ec | 5090 | #ifdef CONFIG_QUOTA |
617ba13b | 5091 | sb->dq_op = &ext4_quota_operations; |
e2b911c5 | 5092 | if (ext4_has_feature_quota(sb)) |
1fa5efe3 | 5093 | sb->s_qcop = &dquot_quotactl_sysfile_ops; |
262b4662 JK |
5094 | else |
5095 | sb->s_qcop = &ext4_qctl_operations; | |
689c958c | 5096 | sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; |
ac27a0ec | 5097 | #endif |
85787090 | 5098 | memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); |
f2fa2ffc | 5099 | |
ac27a0ec | 5100 | INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ |
3b9d4ed2 | 5101 | mutex_init(&sbi->s_orphan_lock); |
ac27a0ec | 5102 | |
aa75f4d3 HS |
5103 | /* Initialize fast commit stuff */ |
5104 | atomic_set(&sbi->s_fc_subtid, 0); | |
5105 | atomic_set(&sbi->s_fc_ineligible_updates, 0); | |
5106 | INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]); | |
5107 | INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]); | |
5108 | INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]); | |
5109 | INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]); | |
5110 | sbi->s_fc_bytes = 0; | |
9b5f6c9b HS |
5111 | ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE); |
5112 | ext4_clear_mount_flag(sb, EXT4_MF_FC_COMMITTING); | |
aa75f4d3 HS |
5113 | spin_lock_init(&sbi->s_fc_lock); |
5114 | memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats)); | |
8016e29f HS |
5115 | sbi->s_fc_replay_state.fc_regions = NULL; |
5116 | sbi->s_fc_replay_state.fc_regions_size = 0; | |
5117 | sbi->s_fc_replay_state.fc_regions_used = 0; | |
5118 | sbi->s_fc_replay_state.fc_regions_valid = 0; | |
5119 | sbi->s_fc_replay_state.fc_modified_inodes = NULL; | |
5120 | sbi->s_fc_replay_state.fc_modified_inodes_size = 0; | |
5121 | sbi->s_fc_replay_state.fc_modified_inodes_used = 0; | |
aa75f4d3 | 5122 | |
ac27a0ec DK |
5123 | sb->s_root = NULL; |
5124 | ||
5125 | needs_recovery = (es->s_last_orphan != 0 || | |
02f310fc | 5126 | ext4_has_feature_orphan_present(sb) || |
e2b911c5 | 5127 | ext4_has_feature_journal_needs_recovery(sb)); |
ac27a0ec | 5128 | |
bc98a42c | 5129 | if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) |
c5e06d10 | 5130 | if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) |
50460fe8 | 5131 | goto failed_mount3a; |
c5e06d10 | 5132 | |
ac27a0ec DK |
5133 | /* |
5134 | * The first inode we look at is the journal inode. Don't try | |
5135 | * root first: it may be modified in the journal! | |
5136 | */ | |
e2b911c5 | 5137 | if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { |
b237e304 | 5138 | err = ext4_load_journal(sb, es, parsed_opts.journal_devnum); |
4753d8a2 | 5139 | if (err) |
50460fe8 | 5140 | goto failed_mount3a; |
bc98a42c | 5141 | } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) && |
e2b911c5 | 5142 | ext4_has_feature_journal_needs_recovery(sb)) { |
b31e1552 ES |
5143 | ext4_msg(sb, KERN_ERR, "required journal recovery " |
5144 | "suppressed and not mounted read-only"); | |
744692dc | 5145 | goto failed_mount_wq; |
ac27a0ec | 5146 | } else { |
1e381f60 DM |
5147 | /* Nojournal mode, all journal mount options are illegal */ |
5148 | if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) { | |
5149 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
5150 | "journal_checksum, fs mounted w/o journal"); | |
5151 | goto failed_mount_wq; | |
5152 | } | |
5153 | if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { | |
5154 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
5155 | "journal_async_commit, fs mounted w/o journal"); | |
5156 | goto failed_mount_wq; | |
5157 | } | |
5158 | if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { | |
5159 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
5160 | "commit=%lu, fs mounted w/o journal", | |
5161 | sbi->s_commit_interval / HZ); | |
5162 | goto failed_mount_wq; | |
5163 | } | |
5164 | if (EXT4_MOUNT_DATA_FLAGS & | |
5165 | (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { | |
5166 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
5167 | "data=, fs mounted w/o journal"); | |
5168 | goto failed_mount_wq; | |
5169 | } | |
50b29d8f | 5170 | sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM; |
1e381f60 | 5171 | clear_opt(sb, JOURNAL_CHECKSUM); |
fd8c37ec | 5172 | clear_opt(sb, DATA_FLAGS); |
995a3ed6 | 5173 | clear_opt2(sb, JOURNAL_FAST_COMMIT); |
0390131b FM |
5174 | sbi->s_journal = NULL; |
5175 | needs_recovery = 0; | |
5176 | goto no_journal; | |
ac27a0ec DK |
5177 | } |
5178 | ||
e2b911c5 | 5179 | if (ext4_has_feature_64bit(sb) && |
eb40a09c JS |
5180 | !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, |
5181 | JBD2_FEATURE_INCOMPAT_64BIT)) { | |
b31e1552 | 5182 | ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); |
744692dc | 5183 | goto failed_mount_wq; |
eb40a09c JS |
5184 | } |
5185 | ||
25ed6e8a DW |
5186 | if (!set_journal_csum_feature_set(sb)) { |
5187 | ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " | |
5188 | "feature set"); | |
5189 | goto failed_mount_wq; | |
d4da6c9c | 5190 | } |
818d276c | 5191 | |
a1e5e465 HS |
5192 | if (test_opt2(sb, JOURNAL_FAST_COMMIT) && |
5193 | !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, | |
5194 | JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) { | |
5195 | ext4_msg(sb, KERN_ERR, | |
5196 | "Failed to set fast commit journal feature"); | |
5197 | goto failed_mount_wq; | |
5198 | } | |
5199 | ||
ac27a0ec DK |
5200 | /* We have now updated the journal if required, so we can |
5201 | * validate the data journaling mode. */ | |
5202 | switch (test_opt(sb, DATA_FLAGS)) { | |
5203 | case 0: | |
5204 | /* No mode set, assume a default based on the journal | |
63f57933 AM |
5205 | * capabilities: ORDERED_DATA if the journal can |
5206 | * cope, else JOURNAL_DATA | |
5207 | */ | |
dab291af | 5208 | if (jbd2_journal_check_available_features |
27f394a7 | 5209 | (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { |
fd8c37ec | 5210 | set_opt(sb, ORDERED_DATA); |
27f394a7 TN |
5211 | sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA; |
5212 | } else { | |
fd8c37ec | 5213 | set_opt(sb, JOURNAL_DATA); |
27f394a7 TN |
5214 | sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA; |
5215 | } | |
ac27a0ec DK |
5216 | break; |
5217 | ||
617ba13b MC |
5218 | case EXT4_MOUNT_ORDERED_DATA: |
5219 | case EXT4_MOUNT_WRITEBACK_DATA: | |
dab291af MC |
5220 | if (!jbd2_journal_check_available_features |
5221 | (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { | |
b31e1552 ES |
5222 | ext4_msg(sb, KERN_ERR, "Journal does not support " |
5223 | "requested data journaling mode"); | |
744692dc | 5224 | goto failed_mount_wq; |
ac27a0ec | 5225 | } |
5a150bde | 5226 | break; |
ac27a0ec DK |
5227 | default: |
5228 | break; | |
5229 | } | |
ab04df78 JK |
5230 | |
5231 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && | |
5232 | test_opt(sb, JOURNAL_ASYNC_COMMIT)) { | |
5233 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
5234 | "journal_async_commit in data=ordered mode"); | |
5235 | goto failed_mount_wq; | |
5236 | } | |
5237 | ||
b237e304 | 5238 | set_task_ioprio(sbi->s_journal->j_task, parsed_opts.journal_ioprio); |
ac27a0ec | 5239 | |
342af94e | 5240 | sbi->s_journal->j_submit_inode_data_buffers = |
afb585a9 | 5241 | ext4_journal_submit_inode_data_buffers; |
342af94e | 5242 | sbi->s_journal->j_finish_inode_data_buffers = |
afb585a9 | 5243 | ext4_journal_finish_inode_data_buffers; |
18aadd47 | 5244 | |
ce7e010a | 5245 | no_journal: |
cdb7ee4c TE |
5246 | if (!test_opt(sb, NO_MBCACHE)) { |
5247 | sbi->s_ea_block_cache = ext4_xattr_create_cache(); | |
5248 | if (!sbi->s_ea_block_cache) { | |
dec214d0 | 5249 | ext4_msg(sb, KERN_ERR, |
cdb7ee4c | 5250 | "Failed to create ea_block_cache"); |
dec214d0 TE |
5251 | goto failed_mount_wq; |
5252 | } | |
cdb7ee4c TE |
5253 | |
5254 | if (ext4_has_feature_ea_inode(sb)) { | |
5255 | sbi->s_ea_inode_cache = ext4_xattr_create_cache(); | |
5256 | if (!sbi->s_ea_inode_cache) { | |
5257 | ext4_msg(sb, KERN_ERR, | |
5258 | "Failed to create ea_inode_cache"); | |
5259 | goto failed_mount_wq; | |
5260 | } | |
5261 | } | |
9c191f70 M |
5262 | } |
5263 | ||
c93d8f88 EB |
5264 | if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) { |
5265 | ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity"); | |
5266 | goto failed_mount_wq; | |
5267 | } | |
5268 | ||
bc98a42c | 5269 | if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) && |
e2b911c5 DW |
5270 | !ext4_has_feature_encrypt(sb)) { |
5271 | ext4_set_feature_encrypt(sb); | |
4392fbc4 | 5272 | ext4_commit_super(sb); |
6ddb2447 TT |
5273 | } |
5274 | ||
952fc18e TT |
5275 | /* |
5276 | * Get the # of file system overhead blocks from the | |
5277 | * superblock if present. | |
5278 | */ | |
5279 | if (es->s_overhead_clusters) | |
5280 | sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); | |
5281 | else { | |
07aa2ea1 LC |
5282 | err = ext4_calculate_overhead(sb); |
5283 | if (err) | |
952fc18e TT |
5284 | goto failed_mount_wq; |
5285 | } | |
5286 | ||
fd89d5f2 TH |
5287 | /* |
5288 | * The maximum number of concurrent works can be high and | |
5289 | * concurrency isn't really necessary. Limit it to 1. | |
5290 | */ | |
2e8fa54e JK |
5291 | EXT4_SB(sb)->rsv_conversion_wq = |
5292 | alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); | |
5293 | if (!EXT4_SB(sb)->rsv_conversion_wq) { | |
5294 | printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); | |
07aa2ea1 | 5295 | ret = -ENOMEM; |
2e8fa54e JK |
5296 | goto failed_mount4; |
5297 | } | |
5298 | ||
ac27a0ec | 5299 | /* |
dab291af | 5300 | * The jbd2_journal_load will have done any necessary log recovery, |
ac27a0ec DK |
5301 | * so we can safely mount the rest of the filesystem now. |
5302 | */ | |
5303 | ||
8a363970 | 5304 | root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL); |
1d1fe1ee | 5305 | if (IS_ERR(root)) { |
b31e1552 | 5306 | ext4_msg(sb, KERN_ERR, "get root inode failed"); |
1d1fe1ee | 5307 | ret = PTR_ERR(root); |
32a9bb57 | 5308 | root = NULL; |
ac27a0ec DK |
5309 | goto failed_mount4; |
5310 | } | |
5311 | if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { | |
b31e1552 | 5312 | ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); |
94bf608a | 5313 | iput(root); |
ac27a0ec DK |
5314 | goto failed_mount4; |
5315 | } | |
b886ee3e | 5316 | |
48fde701 | 5317 | sb->s_root = d_make_root(root); |
1d1fe1ee | 5318 | if (!sb->s_root) { |
b31e1552 | 5319 | ext4_msg(sb, KERN_ERR, "get root dentry failed"); |
1d1fe1ee DH |
5320 | ret = -ENOMEM; |
5321 | goto failed_mount4; | |
5322 | } | |
ac27a0ec | 5323 | |
c89128a0 JK |
5324 | ret = ext4_setup_super(sb, es, sb_rdonly(sb)); |
5325 | if (ret == -EROFS) { | |
1751e8a6 | 5326 | sb->s_flags |= SB_RDONLY; |
c89128a0 JK |
5327 | ret = 0; |
5328 | } else if (ret) | |
5329 | goto failed_mount4a; | |
ef7f3835 | 5330 | |
b5799018 | 5331 | ext4_set_resv_clusters(sb); |
27dd4385 | 5332 | |
0f5bde1d JK |
5333 | if (test_opt(sb, BLOCK_VALIDITY)) { |
5334 | err = ext4_setup_system_zone(sb); | |
5335 | if (err) { | |
5336 | ext4_msg(sb, KERN_ERR, "failed to initialize system " | |
5337 | "zone (%d)", err); | |
5338 | goto failed_mount4a; | |
5339 | } | |
f9ae9cf5 | 5340 | } |
8016e29f | 5341 | ext4_fc_replay_cleanup(sb); |
f9ae9cf5 TT |
5342 | |
5343 | ext4_ext_init(sb); | |
196e402a HS |
5344 | |
5345 | /* | |
5346 | * Enable optimize_scan if number of groups is > threshold. This can be | |
5347 | * turned off by passing "mb_optimize_scan=0". This can also be | |
5348 | * turned on forcefully by passing "mb_optimize_scan=1". | |
5349 | */ | |
5350 | if (parsed_opts.mb_optimize_scan == 1) | |
5351 | set_opt2(sb, MB_OPTIMIZE_SCAN); | |
5352 | else if (parsed_opts.mb_optimize_scan == 0) | |
5353 | clear_opt2(sb, MB_OPTIMIZE_SCAN); | |
5354 | else if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD) | |
5355 | set_opt2(sb, MB_OPTIMIZE_SCAN); | |
5356 | ||
f9ae9cf5 TT |
5357 | err = ext4_mb_init(sb); |
5358 | if (err) { | |
5359 | ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", | |
5360 | err); | |
dcf2d804 | 5361 | goto failed_mount5; |
c2774d84 AK |
5362 | } |
5363 | ||
027f14f5 TT |
5364 | /* |
5365 | * We can only set up the journal commit callback once | |
5366 | * mballoc is initialized | |
5367 | */ | |
5368 | if (sbi->s_journal) | |
5369 | sbi->s_journal->j_commit_callback = | |
5370 | ext4_journal_commit_callback; | |
5371 | ||
d5e03cbb | 5372 | block = ext4_count_free_clusters(sb); |
666245d9 | 5373 | ext4_free_blocks_count_set(sbi->s_es, |
d5e03cbb | 5374 | EXT4_C2B(sbi, block)); |
908c7f19 TH |
5375 | err = percpu_counter_init(&sbi->s_freeclusters_counter, block, |
5376 | GFP_KERNEL); | |
d5e03cbb TT |
5377 | if (!err) { |
5378 | unsigned long freei = ext4_count_free_inodes(sb); | |
5379 | sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); | |
908c7f19 TH |
5380 | err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, |
5381 | GFP_KERNEL); | |
d5e03cbb | 5382 | } |
b2bbb92f JK |
5383 | /* |
5384 | * Update the checksum after updating free space/inode | |
5385 | * counters. Otherwise the superblock can have an incorrect | |
5386 | * checksum in the buffer cache until it is written out and | |
5387 | * e2fsprogs programs trying to open a file system immediately | |
5388 | * after it is mounted can fail. | |
5389 | */ | |
5390 | ext4_superblock_csum_set(sb); | |
d5e03cbb TT |
5391 | if (!err) |
5392 | err = percpu_counter_init(&sbi->s_dirs_counter, | |
908c7f19 | 5393 | ext4_count_dirs(sb), GFP_KERNEL); |
d5e03cbb | 5394 | if (!err) |
908c7f19 TH |
5395 | err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, |
5396 | GFP_KERNEL); | |
efc61345 EW |
5397 | if (!err) |
5398 | err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0, | |
5399 | GFP_KERNEL); | |
c8585c6f | 5400 | if (!err) |
bbd55937 | 5401 | err = percpu_init_rwsem(&sbi->s_writepages_rwsem); |
c8585c6f | 5402 | |
d5e03cbb TT |
5403 | if (err) { |
5404 | ext4_msg(sb, KERN_ERR, "insufficient memory"); | |
5405 | goto failed_mount6; | |
5406 | } | |
5407 | ||
e2b911c5 | 5408 | if (ext4_has_feature_flex_bg(sb)) |
d5e03cbb TT |
5409 | if (!ext4_fill_flex_info(sb)) { |
5410 | ext4_msg(sb, KERN_ERR, | |
5411 | "unable to initialize " | |
5412 | "flex_bg meta info!"); | |
8f6840c4 | 5413 | ret = -ENOMEM; |
d5e03cbb TT |
5414 | goto failed_mount6; |
5415 | } | |
5416 | ||
bfff6873 LC |
5417 | err = ext4_register_li_request(sb, first_not_zeroed); |
5418 | if (err) | |
dcf2d804 | 5419 | goto failed_mount6; |
bfff6873 | 5420 | |
b5799018 | 5421 | err = ext4_register_sysfs(sb); |
dcf2d804 TM |
5422 | if (err) |
5423 | goto failed_mount7; | |
3197ebdb | 5424 | |
02f310fc JK |
5425 | err = ext4_init_orphan_info(sb); |
5426 | if (err) | |
5427 | goto failed_mount8; | |
9b2ff357 JK |
5428 | #ifdef CONFIG_QUOTA |
5429 | /* Enable quota usage during mount. */ | |
bc98a42c | 5430 | if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) { |
9b2ff357 JK |
5431 | err = ext4_enable_quotas(sb); |
5432 | if (err) | |
02f310fc | 5433 | goto failed_mount9; |
9b2ff357 JK |
5434 | } |
5435 | #endif /* CONFIG_QUOTA */ | |
5436 | ||
bc71726c | 5437 | /* |
5438 | * Save the original bdev mapping's wb_err value which could be | |
5439 | * used to detect the metadata async write error. | |
5440 | */ | |
5441 | spin_lock_init(&sbi->s_bdev_wb_lock); | |
9704a322 ZX |
5442 | errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err, |
5443 | &sbi->s_bdev_wb_err); | |
bc71726c | 5444 | sb->s_bdev->bd_super = sb; |
617ba13b MC |
5445 | EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; |
5446 | ext4_orphan_cleanup(sb, es); | |
5447 | EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; | |
0390131b | 5448 | if (needs_recovery) { |
b31e1552 | 5449 | ext4_msg(sb, KERN_INFO, "recovery complete"); |
11215630 JK |
5450 | err = ext4_mark_recovery_complete(sb, es); |
5451 | if (err) | |
02f310fc | 5452 | goto failed_mount9; |
0390131b FM |
5453 | } |
5454 | if (EXT4_SB(sb)->s_journal) { | |
5455 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) | |
5456 | descr = " journalled data mode"; | |
5457 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) | |
5458 | descr = " ordered data mode"; | |
5459 | else | |
5460 | descr = " writeback data mode"; | |
5461 | } else | |
5462 | descr = "out journal"; | |
5463 | ||
79add3a3 LC |
5464 | if (test_opt(sb, DISCARD)) { |
5465 | struct request_queue *q = bdev_get_queue(sb->s_bdev); | |
5466 | if (!blk_queue_discard(q)) | |
5467 | ext4_msg(sb, KERN_WARNING, | |
5468 | "mounting with \"discard\" option, but " | |
5469 | "the device does not support discard"); | |
5470 | } | |
5471 | ||
e294a537 TT |
5472 | if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) |
5473 | ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " | |
ca9b404f | 5474 | "Opts: %.*s%s%s. Quota mode: %s.", descr, |
5aee0f8a TT |
5475 | (int) sizeof(sbi->s_es->s_mount_opts), |
5476 | sbi->s_es->s_mount_opts, | |
ca9b404f RA |
5477 | *sbi->s_es->s_mount_opts ? "; " : "", orig_data, |
5478 | ext4_quota_mode(sb)); | |
ac27a0ec | 5479 | |
66e61a9e TT |
5480 | if (es->s_error_count) |
5481 | mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ | |
ac27a0ec | 5482 | |
efbed4dc TT |
5483 | /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ |
5484 | ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); | |
5485 | ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); | |
5486 | ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); | |
1cf006ed DM |
5487 | atomic_set(&sbi->s_warning_count, 0); |
5488 | atomic_set(&sbi->s_msg_count, 0); | |
efbed4dc | 5489 | |
d4c402d9 | 5490 | kfree(orig_data); |
ac27a0ec DK |
5491 | return 0; |
5492 | ||
617ba13b | 5493 | cantfind_ext4: |
ac27a0ec | 5494 | if (!silent) |
b31e1552 | 5495 | ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); |
ac27a0ec DK |
5496 | goto failed_mount; |
5497 | ||
02f310fc JK |
5498 | failed_mount9: |
5499 | ext4_release_orphan_info(sb); | |
72ba7450 | 5500 | failed_mount8: |
ebd173be | 5501 | ext4_unregister_sysfs(sb); |
cb8d53d2 | 5502 | kobject_put(&sbi->s_kobj); |
dcf2d804 TM |
5503 | failed_mount7: |
5504 | ext4_unregister_li_request(sb); | |
5505 | failed_mount6: | |
f9ae9cf5 | 5506 | ext4_mb_release(sb); |
7c990728 SJS |
5507 | rcu_read_lock(); |
5508 | flex_groups = rcu_dereference(sbi->s_flex_groups); | |
5509 | if (flex_groups) { | |
5510 | for (i = 0; i < sbi->s_flex_groups_allocated; i++) | |
5511 | kvfree(flex_groups[i]); | |
5512 | kvfree(flex_groups); | |
5513 | } | |
5514 | rcu_read_unlock(); | |
d5e03cbb TT |
5515 | percpu_counter_destroy(&sbi->s_freeclusters_counter); |
5516 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
5517 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
5518 | percpu_counter_destroy(&sbi->s_dirtyclusters_counter); | |
efc61345 | 5519 | percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit); |
bbd55937 | 5520 | percpu_free_rwsem(&sbi->s_writepages_rwsem); |
00764937 | 5521 | failed_mount5: |
f9ae9cf5 TT |
5522 | ext4_ext_release(sb); |
5523 | ext4_release_system_zone(sb); | |
5524 | failed_mount4a: | |
94bf608a | 5525 | dput(sb->s_root); |
32a9bb57 | 5526 | sb->s_root = NULL; |
94bf608a | 5527 | failed_mount4: |
b31e1552 | 5528 | ext4_msg(sb, KERN_ERR, "mount failed"); |
2e8fa54e JK |
5529 | if (EXT4_SB(sb)->rsv_conversion_wq) |
5530 | destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); | |
4c0425ff | 5531 | failed_mount_wq: |
50c15df6 CX |
5532 | ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); |
5533 | sbi->s_ea_inode_cache = NULL; | |
5534 | ||
5535 | ext4_xattr_destroy_cache(sbi->s_ea_block_cache); | |
5536 | sbi->s_ea_block_cache = NULL; | |
5537 | ||
0390131b | 5538 | if (sbi->s_journal) { |
bb9464e0 | 5539 | /* flush s_error_work before journal destroy. */ |
5540 | flush_work(&sbi->s_error_work); | |
0390131b FM |
5541 | jbd2_journal_destroy(sbi->s_journal); |
5542 | sbi->s_journal = NULL; | |
5543 | } | |
50460fe8 | 5544 | failed_mount3a: |
d3922a77 | 5545 | ext4_es_unregister_shrinker(sbi); |
eb68d0e2 | 5546 | failed_mount3: |
bb9464e0 | 5547 | /* flush s_error_work before sbi destroy */ |
c92dc856 | 5548 | flush_work(&sbi->s_error_work); |
2a4ae3bc | 5549 | del_timer_sync(&sbi->s_err_report); |
618f0031 | 5550 | ext4_stop_mmpd(sbi); |
ac27a0ec | 5551 | failed_mount2: |
1d0c3924 TT |
5552 | rcu_read_lock(); |
5553 | group_desc = rcu_dereference(sbi->s_group_desc); | |
ac27a0ec | 5554 | for (i = 0; i < db_count; i++) |
1d0c3924 TT |
5555 | brelse(group_desc[i]); |
5556 | kvfree(group_desc); | |
5557 | rcu_read_unlock(); | |
ac27a0ec | 5558 | failed_mount: |
0441984a DW |
5559 | if (sbi->s_chksum_driver) |
5560 | crypto_free_shash(sbi->s_chksum_driver); | |
c83ad55e GKB |
5561 | |
5562 | #ifdef CONFIG_UNICODE | |
f8f4acb6 | 5563 | utf8_unload(sb->s_encoding); |
c83ad55e GKB |
5564 | #endif |
5565 | ||
ac27a0ec | 5566 | #ifdef CONFIG_QUOTA |
a2d4a646 | 5567 | for (i = 0; i < EXT4_MAXQUOTAS; i++) |
0ba33fac | 5568 | kfree(get_qf_name(sb, sbi, i)); |
ac27a0ec | 5569 | #endif |
ac4acb1f | 5570 | fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy); |
afd09b61 | 5571 | /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */ |
ac27a0ec | 5572 | brelse(bh); |
afd09b61 | 5573 | ext4_blkdev_remove(sbi); |
ac27a0ec DK |
5574 | out_fail: |
5575 | sb->s_fs_info = NULL; | |
f6830165 | 5576 | kfree(sbi->s_blockgroup_lock); |
5aee0f8a | 5577 | out_free_base: |
ac27a0ec | 5578 | kfree(sbi); |
d4c402d9 | 5579 | kfree(orig_data); |
5e405595 | 5580 | fs_put_dax(dax_dev); |
07aa2ea1 | 5581 | return err ? err : ret; |
ac27a0ec DK |
5582 | } |
5583 | ||
5584 | /* | |
5585 | * Setup any per-fs journal parameters now. We'll do this both on | |
5586 | * initial mount, once the journal has been initialised but before we've | |
5587 | * done any recovery; and again on any subsequent remount. | |
5588 | */ | |
617ba13b | 5589 | static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) |
ac27a0ec | 5590 | { |
617ba13b | 5591 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 5592 | |
30773840 TT |
5593 | journal->j_commit_interval = sbi->s_commit_interval; |
5594 | journal->j_min_batch_time = sbi->s_min_batch_time; | |
5595 | journal->j_max_batch_time = sbi->s_max_batch_time; | |
6866d7b3 | 5596 | ext4_fc_init(sb, journal); |
ac27a0ec | 5597 | |
a931da6a | 5598 | write_lock(&journal->j_state_lock); |
ac27a0ec | 5599 | if (test_opt(sb, BARRIER)) |
dab291af | 5600 | journal->j_flags |= JBD2_BARRIER; |
ac27a0ec | 5601 | else |
dab291af | 5602 | journal->j_flags &= ~JBD2_BARRIER; |
5bf5683a HK |
5603 | if (test_opt(sb, DATA_ERR_ABORT)) |
5604 | journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; | |
5605 | else | |
5606 | journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; | |
a931da6a | 5607 | write_unlock(&journal->j_state_lock); |
ac27a0ec DK |
5608 | } |
5609 | ||
c6cb7e77 EW |
5610 | static struct inode *ext4_get_journal_inode(struct super_block *sb, |
5611 | unsigned int journal_inum) | |
ac27a0ec DK |
5612 | { |
5613 | struct inode *journal_inode; | |
ac27a0ec | 5614 | |
c6cb7e77 EW |
5615 | /* |
5616 | * Test for the existence of a valid inode on disk. Bad things | |
5617 | * happen if we iget() an unused inode, as the subsequent iput() | |
5618 | * will try to delete it. | |
5619 | */ | |
8a363970 | 5620 | journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL); |
1d1fe1ee | 5621 | if (IS_ERR(journal_inode)) { |
b31e1552 | 5622 | ext4_msg(sb, KERN_ERR, "no journal found"); |
ac27a0ec DK |
5623 | return NULL; |
5624 | } | |
5625 | if (!journal_inode->i_nlink) { | |
5626 | make_bad_inode(journal_inode); | |
5627 | iput(journal_inode); | |
b31e1552 | 5628 | ext4_msg(sb, KERN_ERR, "journal inode is deleted"); |
ac27a0ec DK |
5629 | return NULL; |
5630 | } | |
5631 | ||
e5f8eab8 | 5632 | jbd_debug(2, "Journal inode found at %p: %lld bytes\n", |
ac27a0ec | 5633 | journal_inode, journal_inode->i_size); |
1d1fe1ee | 5634 | if (!S_ISREG(journal_inode->i_mode)) { |
b31e1552 | 5635 | ext4_msg(sb, KERN_ERR, "invalid journal inode"); |
ac27a0ec DK |
5636 | iput(journal_inode); |
5637 | return NULL; | |
5638 | } | |
c6cb7e77 EW |
5639 | return journal_inode; |
5640 | } | |
5641 | ||
5642 | static journal_t *ext4_get_journal(struct super_block *sb, | |
5643 | unsigned int journal_inum) | |
5644 | { | |
5645 | struct inode *journal_inode; | |
5646 | journal_t *journal; | |
5647 | ||
11215630 JK |
5648 | if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) |
5649 | return NULL; | |
c6cb7e77 EW |
5650 | |
5651 | journal_inode = ext4_get_journal_inode(sb, journal_inum); | |
5652 | if (!journal_inode) | |
5653 | return NULL; | |
ac27a0ec | 5654 | |
dab291af | 5655 | journal = jbd2_journal_init_inode(journal_inode); |
ac27a0ec | 5656 | if (!journal) { |
b31e1552 | 5657 | ext4_msg(sb, KERN_ERR, "Could not load journal inode"); |
ac27a0ec DK |
5658 | iput(journal_inode); |
5659 | return NULL; | |
5660 | } | |
5661 | journal->j_private = sb; | |
617ba13b | 5662 | ext4_init_journal_params(sb, journal); |
ac27a0ec DK |
5663 | return journal; |
5664 | } | |
5665 | ||
617ba13b | 5666 | static journal_t *ext4_get_dev_journal(struct super_block *sb, |
ac27a0ec DK |
5667 | dev_t j_dev) |
5668 | { | |
2b2d6d01 | 5669 | struct buffer_head *bh; |
ac27a0ec | 5670 | journal_t *journal; |
617ba13b MC |
5671 | ext4_fsblk_t start; |
5672 | ext4_fsblk_t len; | |
ac27a0ec | 5673 | int hblock, blocksize; |
617ba13b | 5674 | ext4_fsblk_t sb_block; |
ac27a0ec | 5675 | unsigned long offset; |
2b2d6d01 | 5676 | struct ext4_super_block *es; |
ac27a0ec DK |
5677 | struct block_device *bdev; |
5678 | ||
11215630 JK |
5679 | if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) |
5680 | return NULL; | |
0390131b | 5681 | |
b31e1552 | 5682 | bdev = ext4_blkdev_get(j_dev, sb); |
ac27a0ec DK |
5683 | if (bdev == NULL) |
5684 | return NULL; | |
5685 | ||
ac27a0ec | 5686 | blocksize = sb->s_blocksize; |
e1defc4f | 5687 | hblock = bdev_logical_block_size(bdev); |
ac27a0ec | 5688 | if (blocksize < hblock) { |
b31e1552 ES |
5689 | ext4_msg(sb, KERN_ERR, |
5690 | "blocksize too small for journal device"); | |
ac27a0ec DK |
5691 | goto out_bdev; |
5692 | } | |
5693 | ||
617ba13b MC |
5694 | sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; |
5695 | offset = EXT4_MIN_BLOCK_SIZE % blocksize; | |
ac27a0ec DK |
5696 | set_blocksize(bdev, blocksize); |
5697 | if (!(bh = __bread(bdev, sb_block, blocksize))) { | |
b31e1552 ES |
5698 | ext4_msg(sb, KERN_ERR, "couldn't read superblock of " |
5699 | "external journal"); | |
ac27a0ec DK |
5700 | goto out_bdev; |
5701 | } | |
5702 | ||
2716b802 | 5703 | es = (struct ext4_super_block *) (bh->b_data + offset); |
617ba13b | 5704 | if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || |
ac27a0ec | 5705 | !(le32_to_cpu(es->s_feature_incompat) & |
617ba13b | 5706 | EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { |
b31e1552 ES |
5707 | ext4_msg(sb, KERN_ERR, "external journal has " |
5708 | "bad superblock"); | |
ac27a0ec DK |
5709 | brelse(bh); |
5710 | goto out_bdev; | |
5711 | } | |
5712 | ||
df4763be DW |
5713 | if ((le32_to_cpu(es->s_feature_ro_compat) & |
5714 | EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && | |
5715 | es->s_checksum != ext4_superblock_csum(sb, es)) { | |
5716 | ext4_msg(sb, KERN_ERR, "external journal has " | |
5717 | "corrupt superblock"); | |
5718 | brelse(bh); | |
5719 | goto out_bdev; | |
5720 | } | |
5721 | ||
617ba13b | 5722 | if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { |
b31e1552 | 5723 | ext4_msg(sb, KERN_ERR, "journal UUID does not match"); |
ac27a0ec DK |
5724 | brelse(bh); |
5725 | goto out_bdev; | |
5726 | } | |
5727 | ||
bd81d8ee | 5728 | len = ext4_blocks_count(es); |
ac27a0ec DK |
5729 | start = sb_block + 1; |
5730 | brelse(bh); /* we're done with the superblock */ | |
5731 | ||
dab291af | 5732 | journal = jbd2_journal_init_dev(bdev, sb->s_bdev, |
ac27a0ec DK |
5733 | start, len, blocksize); |
5734 | if (!journal) { | |
b31e1552 | 5735 | ext4_msg(sb, KERN_ERR, "failed to create device journal"); |
ac27a0ec DK |
5736 | goto out_bdev; |
5737 | } | |
5738 | journal->j_private = sb; | |
2d069c08 | 5739 | if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) { |
b31e1552 | 5740 | ext4_msg(sb, KERN_ERR, "I/O error on journal device"); |
ac27a0ec DK |
5741 | goto out_journal; |
5742 | } | |
5743 | if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { | |
b31e1552 ES |
5744 | ext4_msg(sb, KERN_ERR, "External journal has more than one " |
5745 | "user (unsupported) - %d", | |
ac27a0ec DK |
5746 | be32_to_cpu(journal->j_superblock->s_nr_users)); |
5747 | goto out_journal; | |
5748 | } | |
ee7ed3aa | 5749 | EXT4_SB(sb)->s_journal_bdev = bdev; |
617ba13b | 5750 | ext4_init_journal_params(sb, journal); |
ac27a0ec | 5751 | return journal; |
0b8e58a1 | 5752 | |
ac27a0ec | 5753 | out_journal: |
dab291af | 5754 | jbd2_journal_destroy(journal); |
ac27a0ec | 5755 | out_bdev: |
617ba13b | 5756 | ext4_blkdev_put(bdev); |
ac27a0ec DK |
5757 | return NULL; |
5758 | } | |
5759 | ||
617ba13b MC |
5760 | static int ext4_load_journal(struct super_block *sb, |
5761 | struct ext4_super_block *es, | |
ac27a0ec DK |
5762 | unsigned long journal_devnum) |
5763 | { | |
5764 | journal_t *journal; | |
5765 | unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); | |
5766 | dev_t journal_dev; | |
5767 | int err = 0; | |
5768 | int really_read_only; | |
273108fa | 5769 | int journal_dev_ro; |
ac27a0ec | 5770 | |
11215630 JK |
5771 | if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) |
5772 | return -EFSCORRUPTED; | |
0390131b | 5773 | |
ac27a0ec DK |
5774 | if (journal_devnum && |
5775 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { | |
b31e1552 ES |
5776 | ext4_msg(sb, KERN_INFO, "external journal device major/minor " |
5777 | "numbers have changed"); | |
ac27a0ec DK |
5778 | journal_dev = new_decode_dev(journal_devnum); |
5779 | } else | |
5780 | journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); | |
5781 | ||
273108fa LC |
5782 | if (journal_inum && journal_dev) { |
5783 | ext4_msg(sb, KERN_ERR, | |
5784 | "filesystem has both journal inode and journal device!"); | |
5785 | return -EINVAL; | |
5786 | } | |
5787 | ||
5788 | if (journal_inum) { | |
5789 | journal = ext4_get_journal(sb, journal_inum); | |
5790 | if (!journal) | |
5791 | return -EINVAL; | |
5792 | } else { | |
5793 | journal = ext4_get_dev_journal(sb, journal_dev); | |
5794 | if (!journal) | |
5795 | return -EINVAL; | |
5796 | } | |
5797 | ||
5798 | journal_dev_ro = bdev_read_only(journal->j_dev); | |
5799 | really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro; | |
5800 | ||
5801 | if (journal_dev_ro && !sb_rdonly(sb)) { | |
5802 | ext4_msg(sb, KERN_ERR, | |
5803 | "journal device read-only, try mounting with '-o ro'"); | |
5804 | err = -EROFS; | |
5805 | goto err_out; | |
5806 | } | |
ac27a0ec DK |
5807 | |
5808 | /* | |
5809 | * Are we loading a blank journal or performing recovery after a | |
5810 | * crash? For recovery, we need to check in advance whether we | |
5811 | * can get read-write access to the device. | |
5812 | */ | |
e2b911c5 | 5813 | if (ext4_has_feature_journal_needs_recovery(sb)) { |
bc98a42c | 5814 | if (sb_rdonly(sb)) { |
b31e1552 ES |
5815 | ext4_msg(sb, KERN_INFO, "INFO: recovery " |
5816 | "required on readonly filesystem"); | |
ac27a0ec | 5817 | if (really_read_only) { |
b31e1552 | 5818 | ext4_msg(sb, KERN_ERR, "write access " |
d98bf8cd SR |
5819 | "unavailable, cannot proceed " |
5820 | "(try mounting with noload)"); | |
273108fa LC |
5821 | err = -EROFS; |
5822 | goto err_out; | |
ac27a0ec | 5823 | } |
b31e1552 ES |
5824 | ext4_msg(sb, KERN_INFO, "write access will " |
5825 | "be enabled during recovery"); | |
ac27a0ec DK |
5826 | } |
5827 | } | |
5828 | ||
90576c0b | 5829 | if (!(journal->j_flags & JBD2_BARRIER)) |
b31e1552 | 5830 | ext4_msg(sb, KERN_INFO, "barriers disabled"); |
4776004f | 5831 | |
e2b911c5 | 5832 | if (!ext4_has_feature_journal_needs_recovery(sb)) |
dab291af | 5833 | err = jbd2_journal_wipe(journal, !really_read_only); |
1c13d5c0 TT |
5834 | if (!err) { |
5835 | char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); | |
5836 | if (save) | |
5837 | memcpy(save, ((char *) es) + | |
5838 | EXT4_S_ERR_START, EXT4_S_ERR_LEN); | |
dab291af | 5839 | err = jbd2_journal_load(journal); |
1c13d5c0 TT |
5840 | if (save) |
5841 | memcpy(((char *) es) + EXT4_S_ERR_START, | |
5842 | save, EXT4_S_ERR_LEN); | |
5843 | kfree(save); | |
5844 | } | |
ac27a0ec DK |
5845 | |
5846 | if (err) { | |
b31e1552 | 5847 | ext4_msg(sb, KERN_ERR, "error loading journal"); |
273108fa | 5848 | goto err_out; |
ac27a0ec DK |
5849 | } |
5850 | ||
617ba13b | 5851 | EXT4_SB(sb)->s_journal = journal; |
11215630 JK |
5852 | err = ext4_clear_journal_err(sb, es); |
5853 | if (err) { | |
5854 | EXT4_SB(sb)->s_journal = NULL; | |
5855 | jbd2_journal_destroy(journal); | |
5856 | return err; | |
5857 | } | |
ac27a0ec | 5858 | |
c41303ce | 5859 | if (!really_read_only && journal_devnum && |
ac27a0ec DK |
5860 | journal_devnum != le32_to_cpu(es->s_journal_dev)) { |
5861 | es->s_journal_dev = cpu_to_le32(journal_devnum); | |
ac27a0ec DK |
5862 | |
5863 | /* Make sure we flush the recovery flag to disk. */ | |
4392fbc4 | 5864 | ext4_commit_super(sb); |
ac27a0ec DK |
5865 | } |
5866 | ||
5867 | return 0; | |
273108fa LC |
5868 | |
5869 | err_out: | |
5870 | jbd2_journal_destroy(journal); | |
5871 | return err; | |
ac27a0ec DK |
5872 | } |
5873 | ||
2d01ddc8 JK |
5874 | /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */ |
5875 | static void ext4_update_super(struct super_block *sb) | |
ac27a0ec | 5876 | { |
c92dc856 | 5877 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
e92ad03f JK |
5878 | struct ext4_super_block *es = sbi->s_es; |
5879 | struct buffer_head *sbh = sbi->s_sbh; | |
a17712c8 | 5880 | |
05c2c00f | 5881 | lock_buffer(sbh); |
71290b36 TT |
5882 | /* |
5883 | * If the file system is mounted read-only, don't update the | |
5884 | * superblock write time. This avoids updating the superblock | |
5885 | * write time when we are mounting the root file system | |
5886 | * read/only but we need to replay the journal; at that point, | |
5887 | * for people who are east of GMT and who make their clock | |
5888 | * tick in localtime for Windows bug-for-bug compatibility, | |
5889 | * the clock is set in the future, and this will cause e2fsck | |
5890 | * to complain and force a full file system check. | |
5891 | */ | |
1751e8a6 | 5892 | if (!(sb->s_flags & SB_RDONLY)) |
6a0678a7 | 5893 | ext4_update_tstamp(es, s_wtime); |
8446fe92 | 5894 | es->s_kbytes_written = |
0bc9bc1d | 5895 | cpu_to_le64(sbi->s_kbytes_written + |
8446fe92 | 5896 | ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) - |
0bc9bc1d | 5897 | sbi->s_sectors_written_start) >> 1)); |
e92ad03f | 5898 | if (percpu_counter_initialized(&sbi->s_freeclusters_counter)) |
d5e03cbb | 5899 | ext4_free_blocks_count_set(es, |
e92ad03f JK |
5900 | EXT4_C2B(sbi, percpu_counter_sum_positive( |
5901 | &sbi->s_freeclusters_counter))); | |
5902 | if (percpu_counter_initialized(&sbi->s_freeinodes_counter)) | |
d5e03cbb TT |
5903 | es->s_free_inodes_count = |
5904 | cpu_to_le32(percpu_counter_sum_positive( | |
e92ad03f | 5905 | &sbi->s_freeinodes_counter)); |
c92dc856 JK |
5906 | /* Copy error information to the on-disk superblock */ |
5907 | spin_lock(&sbi->s_error_lock); | |
5908 | if (sbi->s_add_error_count > 0) { | |
5909 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
5910 | if (!es->s_first_error_time && !es->s_first_error_time_hi) { | |
5911 | __ext4_update_tstamp(&es->s_first_error_time, | |
5912 | &es->s_first_error_time_hi, | |
5913 | sbi->s_first_error_time); | |
5914 | strncpy(es->s_first_error_func, sbi->s_first_error_func, | |
5915 | sizeof(es->s_first_error_func)); | |
5916 | es->s_first_error_line = | |
5917 | cpu_to_le32(sbi->s_first_error_line); | |
5918 | es->s_first_error_ino = | |
5919 | cpu_to_le32(sbi->s_first_error_ino); | |
5920 | es->s_first_error_block = | |
5921 | cpu_to_le64(sbi->s_first_error_block); | |
5922 | es->s_first_error_errcode = | |
5923 | ext4_errno_to_code(sbi->s_first_error_code); | |
5924 | } | |
5925 | __ext4_update_tstamp(&es->s_last_error_time, | |
5926 | &es->s_last_error_time_hi, | |
5927 | sbi->s_last_error_time); | |
5928 | strncpy(es->s_last_error_func, sbi->s_last_error_func, | |
5929 | sizeof(es->s_last_error_func)); | |
5930 | es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line); | |
5931 | es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino); | |
5932 | es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block); | |
5933 | es->s_last_error_errcode = | |
5934 | ext4_errno_to_code(sbi->s_last_error_code); | |
5935 | /* | |
5936 | * Start the daily error reporting function if it hasn't been | |
5937 | * started already | |
5938 | */ | |
5939 | if (!es->s_error_count) | |
5940 | mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); | |
5941 | le32_add_cpu(&es->s_error_count, sbi->s_add_error_count); | |
5942 | sbi->s_add_error_count = 0; | |
5943 | } | |
5944 | spin_unlock(&sbi->s_error_lock); | |
5945 | ||
06db49e6 | 5946 | ext4_superblock_csum_set(sb); |
2d01ddc8 JK |
5947 | unlock_buffer(sbh); |
5948 | } | |
5949 | ||
5950 | static int ext4_commit_super(struct super_block *sb) | |
5951 | { | |
5952 | struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; | |
5953 | int error = 0; | |
5954 | ||
f88f1466 FC |
5955 | if (!sbh) |
5956 | return -EINVAL; | |
5957 | if (block_device_ejected(sb)) | |
5958 | return -ENODEV; | |
2d01ddc8 JK |
5959 | |
5960 | ext4_update_super(sb); | |
5961 | ||
e8680786 | 5962 | if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { |
4743f839 PKS |
5963 | /* |
5964 | * Oh, dear. A previous attempt to write the | |
5965 | * superblock failed. This could happen because the | |
5966 | * USB device was yanked out. Or it could happen to | |
5967 | * be a transient write error and maybe the block will | |
5968 | * be remapped. Nothing we can do but to retry the | |
5969 | * write and hope for the best. | |
5970 | */ | |
5971 | ext4_msg(sb, KERN_ERR, "previous I/O error to " | |
5972 | "superblock detected"); | |
5973 | clear_buffer_write_io_error(sbh); | |
5974 | set_buffer_uptodate(sbh); | |
5975 | } | |
2d01ddc8 | 5976 | BUFFER_TRACE(sbh, "marking dirty"); |
ac27a0ec | 5977 | mark_buffer_dirty(sbh); |
4392fbc4 JK |
5978 | error = __sync_dirty_buffer(sbh, |
5979 | REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0)); | |
5980 | if (buffer_write_io_error(sbh)) { | |
5981 | ext4_msg(sb, KERN_ERR, "I/O error while writing " | |
5982 | "superblock"); | |
5983 | clear_buffer_write_io_error(sbh); | |
5984 | set_buffer_uptodate(sbh); | |
914258bf | 5985 | } |
c4be0c1d | 5986 | return error; |
ac27a0ec DK |
5987 | } |
5988 | ||
ac27a0ec DK |
5989 | /* |
5990 | * Have we just finished recovery? If so, and if we are mounting (or | |
5991 | * remounting) the filesystem readonly, then we will end up with a | |
5992 | * consistent fs on disk. Record that fact. | |
5993 | */ | |
11215630 JK |
5994 | static int ext4_mark_recovery_complete(struct super_block *sb, |
5995 | struct ext4_super_block *es) | |
ac27a0ec | 5996 | { |
11215630 | 5997 | int err; |
617ba13b | 5998 | journal_t *journal = EXT4_SB(sb)->s_journal; |
ac27a0ec | 5999 | |
e2b911c5 | 6000 | if (!ext4_has_feature_journal(sb)) { |
11215630 JK |
6001 | if (journal != NULL) { |
6002 | ext4_error(sb, "Journal got removed while the fs was " | |
6003 | "mounted!"); | |
6004 | return -EFSCORRUPTED; | |
6005 | } | |
6006 | return 0; | |
0390131b | 6007 | } |
dab291af | 6008 | jbd2_journal_lock_updates(journal); |
01d5d965 | 6009 | err = jbd2_journal_flush(journal, 0); |
11215630 | 6010 | if (err < 0) |
7ffe1ea8 HK |
6011 | goto out; |
6012 | ||
02f310fc JK |
6013 | if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) || |
6014 | ext4_has_feature_orphan_present(sb))) { | |
6015 | if (!ext4_orphan_file_empty(sb)) { | |
6016 | ext4_error(sb, "Orphan file not empty on read-only fs."); | |
6017 | err = -EFSCORRUPTED; | |
6018 | goto out; | |
6019 | } | |
e2b911c5 | 6020 | ext4_clear_feature_journal_needs_recovery(sb); |
02f310fc | 6021 | ext4_clear_feature_orphan_present(sb); |
4392fbc4 | 6022 | ext4_commit_super(sb); |
ac27a0ec | 6023 | } |
7ffe1ea8 | 6024 | out: |
dab291af | 6025 | jbd2_journal_unlock_updates(journal); |
11215630 | 6026 | return err; |
ac27a0ec DK |
6027 | } |
6028 | ||
6029 | /* | |
6030 | * If we are mounting (or read-write remounting) a filesystem whose journal | |
6031 | * has recorded an error from a previous lifetime, move that error to the | |
6032 | * main filesystem now. | |
6033 | */ | |
11215630 | 6034 | static int ext4_clear_journal_err(struct super_block *sb, |
2b2d6d01 | 6035 | struct ext4_super_block *es) |
ac27a0ec DK |
6036 | { |
6037 | journal_t *journal; | |
6038 | int j_errno; | |
6039 | const char *errstr; | |
6040 | ||
11215630 JK |
6041 | if (!ext4_has_feature_journal(sb)) { |
6042 | ext4_error(sb, "Journal got removed while the fs was mounted!"); | |
6043 | return -EFSCORRUPTED; | |
6044 | } | |
0390131b | 6045 | |
617ba13b | 6046 | journal = EXT4_SB(sb)->s_journal; |
ac27a0ec DK |
6047 | |
6048 | /* | |
6049 | * Now check for any error status which may have been recorded in the | |
617ba13b | 6050 | * journal by a prior ext4_error() or ext4_abort() |
ac27a0ec DK |
6051 | */ |
6052 | ||
dab291af | 6053 | j_errno = jbd2_journal_errno(journal); |
ac27a0ec DK |
6054 | if (j_errno) { |
6055 | char nbuf[16]; | |
6056 | ||
617ba13b | 6057 | errstr = ext4_decode_error(sb, j_errno, nbuf); |
12062ddd | 6058 | ext4_warning(sb, "Filesystem error recorded " |
ac27a0ec | 6059 | "from previous mount: %s", errstr); |
12062ddd | 6060 | ext4_warning(sb, "Marking fs in need of filesystem check."); |
ac27a0ec | 6061 | |
617ba13b MC |
6062 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; |
6063 | es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
4392fbc4 | 6064 | ext4_commit_super(sb); |
ac27a0ec | 6065 | |
dab291af | 6066 | jbd2_journal_clear_err(journal); |
d796c52e | 6067 | jbd2_journal_update_sb_errno(journal); |
ac27a0ec | 6068 | } |
11215630 | 6069 | return 0; |
ac27a0ec DK |
6070 | } |
6071 | ||
6072 | /* | |
6073 | * Force the running and committing transactions to commit, | |
6074 | * and wait on the commit. | |
6075 | */ | |
617ba13b | 6076 | int ext4_force_commit(struct super_block *sb) |
ac27a0ec DK |
6077 | { |
6078 | journal_t *journal; | |
ac27a0ec | 6079 | |
bc98a42c | 6080 | if (sb_rdonly(sb)) |
ac27a0ec DK |
6081 | return 0; |
6082 | ||
617ba13b | 6083 | journal = EXT4_SB(sb)->s_journal; |
b1deefc9 | 6084 | return ext4_journal_force_commit(journal); |
ac27a0ec DK |
6085 | } |
6086 | ||
617ba13b | 6087 | static int ext4_sync_fs(struct super_block *sb, int wait) |
ac27a0ec | 6088 | { |
14ce0cb4 | 6089 | int ret = 0; |
9eddacf9 | 6090 | tid_t target; |
06a407f1 | 6091 | bool needs_barrier = false; |
8d5d02e6 | 6092 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
ac27a0ec | 6093 | |
49598e04 | 6094 | if (unlikely(ext4_forced_shutdown(sbi))) |
0db1ff22 TT |
6095 | return 0; |
6096 | ||
9bffad1e | 6097 | trace_ext4_sync_fs(sb, wait); |
2e8fa54e | 6098 | flush_workqueue(sbi->rsv_conversion_wq); |
a1177825 JK |
6099 | /* |
6100 | * Writeback quota in non-journalled quota case - journalled quota has | |
6101 | * no dirty dquots | |
6102 | */ | |
6103 | dquot_writeback_dquots(sb, -1); | |
06a407f1 DM |
6104 | /* |
6105 | * Data writeback is possible w/o journal transaction, so barrier must | |
6106 | * being sent at the end of the function. But we can skip it if | |
6107 | * transaction_commit will do it for us. | |
6108 | */ | |
bda32530 TT |
6109 | if (sbi->s_journal) { |
6110 | target = jbd2_get_latest_transaction(sbi->s_journal); | |
6111 | if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && | |
6112 | !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) | |
6113 | needs_barrier = true; | |
6114 | ||
6115 | if (jbd2_journal_start_commit(sbi->s_journal, &target)) { | |
6116 | if (wait) | |
6117 | ret = jbd2_log_wait_commit(sbi->s_journal, | |
6118 | target); | |
6119 | } | |
6120 | } else if (wait && test_opt(sb, BARRIER)) | |
06a407f1 | 6121 | needs_barrier = true; |
06a407f1 DM |
6122 | if (needs_barrier) { |
6123 | int err; | |
c6bf3f0e | 6124 | err = blkdev_issue_flush(sb->s_bdev); |
06a407f1 DM |
6125 | if (!ret) |
6126 | ret = err; | |
0390131b | 6127 | } |
06a407f1 DM |
6128 | |
6129 | return ret; | |
6130 | } | |
6131 | ||
ac27a0ec DK |
6132 | /* |
6133 | * LVM calls this function before a (read-only) snapshot is created. This | |
6134 | * gives us a chance to flush the journal completely and mark the fs clean. | |
be4f27d3 YY |
6135 | * |
6136 | * Note that only this function cannot bring a filesystem to be in a clean | |
8e8ad8a5 JK |
6137 | * state independently. It relies on upper layer to stop all data & metadata |
6138 | * modifications. | |
ac27a0ec | 6139 | */ |
c4be0c1d | 6140 | static int ext4_freeze(struct super_block *sb) |
ac27a0ec | 6141 | { |
c4be0c1d TS |
6142 | int error = 0; |
6143 | journal_t *journal; | |
ac27a0ec | 6144 | |
bc98a42c | 6145 | if (sb_rdonly(sb)) |
9ca92389 | 6146 | return 0; |
ac27a0ec | 6147 | |
9ca92389 | 6148 | journal = EXT4_SB(sb)->s_journal; |
7ffe1ea8 | 6149 | |
bb044576 TT |
6150 | if (journal) { |
6151 | /* Now we set up the journal barrier. */ | |
6152 | jbd2_journal_lock_updates(journal); | |
ac27a0ec | 6153 | |
bb044576 TT |
6154 | /* |
6155 | * Don't clear the needs_recovery flag if we failed to | |
6156 | * flush the journal. | |
6157 | */ | |
01d5d965 | 6158 | error = jbd2_journal_flush(journal, 0); |
bb044576 TT |
6159 | if (error < 0) |
6160 | goto out; | |
c642dc9e ES |
6161 | |
6162 | /* Journal blocked and flushed, clear needs_recovery flag. */ | |
e2b911c5 | 6163 | ext4_clear_feature_journal_needs_recovery(sb); |
02f310fc JK |
6164 | if (ext4_orphan_file_empty(sb)) |
6165 | ext4_clear_feature_orphan_present(sb); | |
bb044576 | 6166 | } |
9ca92389 | 6167 | |
4392fbc4 | 6168 | error = ext4_commit_super(sb); |
6b0310fb | 6169 | out: |
bb044576 TT |
6170 | if (journal) |
6171 | /* we rely on upper layer to stop further updates */ | |
6172 | jbd2_journal_unlock_updates(journal); | |
6b0310fb | 6173 | return error; |
ac27a0ec DK |
6174 | } |
6175 | ||
6176 | /* | |
6177 | * Called by LVM after the snapshot is done. We need to reset the RECOVER | |
6178 | * flag here, even though the filesystem is not technically dirty yet. | |
6179 | */ | |
c4be0c1d | 6180 | static int ext4_unfreeze(struct super_block *sb) |
ac27a0ec | 6181 | { |
bc98a42c | 6182 | if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb))) |
9ca92389 TT |
6183 | return 0; |
6184 | ||
c642dc9e ES |
6185 | if (EXT4_SB(sb)->s_journal) { |
6186 | /* Reset the needs_recovery flag before the fs is unlocked. */ | |
e2b911c5 | 6187 | ext4_set_feature_journal_needs_recovery(sb); |
02f310fc JK |
6188 | if (ext4_has_feature_orphan_file(sb)) |
6189 | ext4_set_feature_orphan_present(sb); | |
c642dc9e ES |
6190 | } |
6191 | ||
4392fbc4 | 6192 | ext4_commit_super(sb); |
c4be0c1d | 6193 | return 0; |
ac27a0ec DK |
6194 | } |
6195 | ||
673c6100 TT |
6196 | /* |
6197 | * Structure to save mount options for ext4_remount's benefit | |
6198 | */ | |
6199 | struct ext4_mount_options { | |
6200 | unsigned long s_mount_opt; | |
a2595b8a | 6201 | unsigned long s_mount_opt2; |
08cefc7a EB |
6202 | kuid_t s_resuid; |
6203 | kgid_t s_resgid; | |
673c6100 TT |
6204 | unsigned long s_commit_interval; |
6205 | u32 s_min_batch_time, s_max_batch_time; | |
6206 | #ifdef CONFIG_QUOTA | |
6207 | int s_jquota_fmt; | |
a2d4a646 | 6208 | char *s_qf_names[EXT4_MAXQUOTAS]; |
673c6100 TT |
6209 | #endif |
6210 | }; | |
6211 | ||
2b2d6d01 | 6212 | static int ext4_remount(struct super_block *sb, int *flags, char *data) |
ac27a0ec | 6213 | { |
2b2d6d01 | 6214 | struct ext4_super_block *es; |
617ba13b | 6215 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
f25391eb | 6216 | unsigned long old_sb_flags, vfs_flags; |
617ba13b | 6217 | struct ext4_mount_options old_opts; |
8a266467 | 6218 | ext4_group_t g; |
c5e06d10 | 6219 | int err = 0; |
ac27a0ec | 6220 | #ifdef CONFIG_QUOTA |
3bbef91b | 6221 | int enable_quota = 0; |
03dafb5f | 6222 | int i, j; |
33458eab | 6223 | char *to_free[EXT4_MAXQUOTAS]; |
ac27a0ec | 6224 | #endif |
d4c402d9 | 6225 | char *orig_data = kstrdup(data, GFP_KERNEL); |
461c3af0 | 6226 | struct ext4_fs_context parsed_opts; |
b237e304 HS |
6227 | |
6228 | parsed_opts.journal_ioprio = DEFAULT_JOURNAL_IOPRIO; | |
6229 | parsed_opts.journal_devnum = 0; | |
ac27a0ec | 6230 | |
21ac738e CX |
6231 | if (data && !orig_data) |
6232 | return -ENOMEM; | |
6233 | ||
ac27a0ec DK |
6234 | /* Store the original options */ |
6235 | old_sb_flags = sb->s_flags; | |
6236 | old_opts.s_mount_opt = sbi->s_mount_opt; | |
a2595b8a | 6237 | old_opts.s_mount_opt2 = sbi->s_mount_opt2; |
ac27a0ec DK |
6238 | old_opts.s_resuid = sbi->s_resuid; |
6239 | old_opts.s_resgid = sbi->s_resgid; | |
6240 | old_opts.s_commit_interval = sbi->s_commit_interval; | |
30773840 TT |
6241 | old_opts.s_min_batch_time = sbi->s_min_batch_time; |
6242 | old_opts.s_max_batch_time = sbi->s_max_batch_time; | |
ac27a0ec DK |
6243 | #ifdef CONFIG_QUOTA |
6244 | old_opts.s_jquota_fmt = sbi->s_jquota_fmt; | |
a2d4a646 | 6245 | for (i = 0; i < EXT4_MAXQUOTAS; i++) |
03dafb5f | 6246 | if (sbi->s_qf_names[i]) { |
33458eab TT |
6247 | char *qf_name = get_qf_name(sb, sbi, i); |
6248 | ||
6249 | old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); | |
03dafb5f CG |
6250 | if (!old_opts.s_qf_names[i]) { |
6251 | for (j = 0; j < i; j++) | |
6252 | kfree(old_opts.s_qf_names[j]); | |
3e36a163 | 6253 | kfree(orig_data); |
03dafb5f CG |
6254 | return -ENOMEM; |
6255 | } | |
6256 | } else | |
6257 | old_opts.s_qf_names[i] = NULL; | |
ac27a0ec | 6258 | #endif |
b3881f74 | 6259 | if (sbi->s_journal && sbi->s_journal->j_task->io_context) |
b237e304 HS |
6260 | parsed_opts.journal_ioprio = |
6261 | sbi->s_journal->j_task->io_context->ioprio; | |
ac27a0ec | 6262 | |
f25391eb LC |
6263 | /* |
6264 | * Some options can be enabled by ext4 and/or by VFS mount flag | |
6265 | * either way we need to make sure it matches in both *flags and | |
6266 | * s_flags. Copy those selected flags from *flags to s_flags | |
6267 | */ | |
6268 | vfs_flags = SB_LAZYTIME | SB_I_VERSION; | |
6269 | sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags); | |
6270 | ||
b237e304 | 6271 | if (!parse_options(data, sb, &parsed_opts, 1)) { |
ac27a0ec DK |
6272 | err = -EINVAL; |
6273 | goto restore_opts; | |
6274 | } | |
6275 | ||
6b992ff2 | 6276 | if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ |
c6d3d56d DW |
6277 | test_opt(sb, JOURNAL_CHECKSUM)) { |
6278 | ext4_msg(sb, KERN_ERR, "changing journal_checksum " | |
2d5b86e0 ES |
6279 | "during remount not supported; ignoring"); |
6280 | sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; | |
6b992ff2 DW |
6281 | } |
6282 | ||
6ae6514b PS |
6283 | if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { |
6284 | if (test_opt2(sb, EXPLICIT_DELALLOC)) { | |
6285 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
6286 | "both data=journal and delalloc"); | |
6287 | err = -EINVAL; | |
6288 | goto restore_opts; | |
6289 | } | |
6290 | if (test_opt(sb, DIOREAD_NOLOCK)) { | |
6291 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
6292 | "both data=journal and dioread_nolock"); | |
6293 | err = -EINVAL; | |
6294 | goto restore_opts; | |
6295 | } | |
ab04df78 JK |
6296 | } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) { |
6297 | if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { | |
6298 | ext4_msg(sb, KERN_ERR, "can't mount with " | |
6299 | "journal_async_commit in data=ordered mode"); | |
6300 | err = -EINVAL; | |
6301 | goto restore_opts; | |
6302 | } | |
923ae0ff RZ |
6303 | } |
6304 | ||
cdb7ee4c TE |
6305 | if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) { |
6306 | ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount"); | |
6307 | err = -EINVAL; | |
6308 | goto restore_opts; | |
6309 | } | |
6310 | ||
9b5f6c9b | 6311 | if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) |
124e7c61 | 6312 | ext4_abort(sb, ESHUTDOWN, "Abort forced by user"); |
ac27a0ec | 6313 | |
1751e8a6 LT |
6314 | sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | |
6315 | (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); | |
ac27a0ec DK |
6316 | |
6317 | es = sbi->s_es; | |
6318 | ||
b3881f74 | 6319 | if (sbi->s_journal) { |
0390131b | 6320 | ext4_init_journal_params(sb, sbi->s_journal); |
b237e304 | 6321 | set_task_ioprio(sbi->s_journal->j_task, parsed_opts.journal_ioprio); |
b3881f74 | 6322 | } |
ac27a0ec | 6323 | |
c92dc856 JK |
6324 | /* Flush outstanding errors before changing fs state */ |
6325 | flush_work(&sbi->s_error_work); | |
6326 | ||
1751e8a6 | 6327 | if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) { |
9b5f6c9b | 6328 | if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) { |
ac27a0ec DK |
6329 | err = -EROFS; |
6330 | goto restore_opts; | |
6331 | } | |
6332 | ||
1751e8a6 | 6333 | if (*flags & SB_RDONLY) { |
38c03b34 TT |
6334 | err = sync_filesystem(sb); |
6335 | if (err < 0) | |
6336 | goto restore_opts; | |
0f0dd62f CH |
6337 | err = dquot_suspend(sb, -1); |
6338 | if (err < 0) | |
c79d967d | 6339 | goto restore_opts; |
c79d967d | 6340 | |
ac27a0ec DK |
6341 | /* |
6342 | * First of all, the unconditional stuff we have to do | |
6343 | * to disable replay of the journal when we next remount | |
6344 | */ | |
1751e8a6 | 6345 | sb->s_flags |= SB_RDONLY; |
ac27a0ec DK |
6346 | |
6347 | /* | |
6348 | * OK, test if we are remounting a valid rw partition | |
6349 | * readonly, and if so set the rdonly flag and then | |
6350 | * mark the partition as valid again. | |
6351 | */ | |
617ba13b MC |
6352 | if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && |
6353 | (sbi->s_mount_state & EXT4_VALID_FS)) | |
ac27a0ec DK |
6354 | es->s_state = cpu_to_le16(sbi->s_mount_state); |
6355 | ||
11215630 JK |
6356 | if (sbi->s_journal) { |
6357 | /* | |
6358 | * We let remount-ro finish even if marking fs | |
6359 | * as clean failed... | |
6360 | */ | |
0390131b | 6361 | ext4_mark_recovery_complete(sb, es); |
11215630 | 6362 | } |
ac27a0ec | 6363 | } else { |
a13fb1a4 | 6364 | /* Make sure we can mount this feature set readwrite */ |
e2b911c5 | 6365 | if (ext4_has_feature_readonly(sb) || |
2cb5cc8b | 6366 | !ext4_feature_set_ok(sb, 0)) { |
ac27a0ec DK |
6367 | err = -EROFS; |
6368 | goto restore_opts; | |
6369 | } | |
8a266467 TT |
6370 | /* |
6371 | * Make sure the group descriptor checksums | |
0b8e58a1 | 6372 | * are sane. If they aren't, refuse to remount r/w. |
8a266467 TT |
6373 | */ |
6374 | for (g = 0; g < sbi->s_groups_count; g++) { | |
6375 | struct ext4_group_desc *gdp = | |
6376 | ext4_get_group_desc(sb, g, NULL); | |
6377 | ||
feb0ab32 | 6378 | if (!ext4_group_desc_csum_verify(sb, g, gdp)) { |
b31e1552 ES |
6379 | ext4_msg(sb, KERN_ERR, |
6380 | "ext4_remount: Checksum for group %u failed (%u!=%u)", | |
e2b911c5 | 6381 | g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), |
8a266467 | 6382 | le16_to_cpu(gdp->bg_checksum)); |
6a797d27 | 6383 | err = -EFSBADCRC; |
8a266467 TT |
6384 | goto restore_opts; |
6385 | } | |
6386 | } | |
6387 | ||
ead6596b ES |
6388 | /* |
6389 | * If we have an unprocessed orphan list hanging | |
6390 | * around from a previously readonly bdev mount, | |
6391 | * require a full umount/remount for now. | |
6392 | */ | |
02f310fc | 6393 | if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) { |
b31e1552 | 6394 | ext4_msg(sb, KERN_WARNING, "Couldn't " |
ead6596b ES |
6395 | "remount RDWR because of unprocessed " |
6396 | "orphan inode list. Please " | |
b31e1552 | 6397 | "umount/remount instead"); |
ead6596b ES |
6398 | err = -EINVAL; |
6399 | goto restore_opts; | |
6400 | } | |
6401 | ||
ac27a0ec DK |
6402 | /* |
6403 | * Mounting a RDONLY partition read-write, so reread | |
6404 | * and store the current valid flag. (It may have | |
6405 | * been changed by e2fsck since we originally mounted | |
6406 | * the partition.) | |
6407 | */ | |
11215630 JK |
6408 | if (sbi->s_journal) { |
6409 | err = ext4_clear_journal_err(sb, es); | |
6410 | if (err) | |
6411 | goto restore_opts; | |
6412 | } | |
ac27a0ec | 6413 | sbi->s_mount_state = le16_to_cpu(es->s_state); |
c89128a0 JK |
6414 | |
6415 | err = ext4_setup_super(sb, es, 0); | |
6416 | if (err) | |
6417 | goto restore_opts; | |
6418 | ||
6419 | sb->s_flags &= ~SB_RDONLY; | |
e2b911c5 | 6420 | if (ext4_has_feature_mmp(sb)) |
c5e06d10 JL |
6421 | if (ext4_multi_mount_protect(sb, |
6422 | le64_to_cpu(es->s_mmp_block))) { | |
6423 | err = -EROFS; | |
6424 | goto restore_opts; | |
6425 | } | |
3bbef91b | 6426 | #ifdef CONFIG_QUOTA |
c79d967d | 6427 | enable_quota = 1; |
3bbef91b | 6428 | #endif |
ac27a0ec DK |
6429 | } |
6430 | } | |
bfff6873 LC |
6431 | |
6432 | /* | |
6433 | * Reinitialize lazy itable initialization thread based on | |
6434 | * current settings | |
6435 | */ | |
bc98a42c | 6436 | if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE)) |
bfff6873 LC |
6437 | ext4_unregister_li_request(sb); |
6438 | else { | |
6439 | ext4_group_t first_not_zeroed; | |
6440 | first_not_zeroed = ext4_has_uninit_itable(sb); | |
6441 | ext4_register_li_request(sb, first_not_zeroed); | |
6442 | } | |
6443 | ||
0f5bde1d JK |
6444 | /* |
6445 | * Handle creation of system zone data early because it can fail. | |
6446 | * Releasing of existing data is done when we are sure remount will | |
6447 | * succeed. | |
6448 | */ | |
dd0db94f | 6449 | if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) { |
0f5bde1d JK |
6450 | err = ext4_setup_system_zone(sb); |
6451 | if (err) | |
6452 | goto restore_opts; | |
6453 | } | |
d176b1f6 | 6454 | |
c89128a0 | 6455 | if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) { |
4392fbc4 | 6456 | err = ext4_commit_super(sb); |
c89128a0 JK |
6457 | if (err) |
6458 | goto restore_opts; | |
6459 | } | |
0390131b | 6460 | |
ac27a0ec DK |
6461 | #ifdef CONFIG_QUOTA |
6462 | /* Release old quota file names */ | |
a2d4a646 | 6463 | for (i = 0; i < EXT4_MAXQUOTAS; i++) |
03dafb5f | 6464 | kfree(old_opts.s_qf_names[i]); |
7c319d32 AK |
6465 | if (enable_quota) { |
6466 | if (sb_any_quota_suspended(sb)) | |
6467 | dquot_resume(sb, -1); | |
e2b911c5 | 6468 | else if (ext4_has_feature_quota(sb)) { |
7c319d32 | 6469 | err = ext4_enable_quotas(sb); |
07724f98 | 6470 | if (err) |
7c319d32 | 6471 | goto restore_opts; |
7c319d32 AK |
6472 | } |
6473 | } | |
ac27a0ec | 6474 | #endif |
dd0db94f | 6475 | if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) |
0f5bde1d | 6476 | ext4_release_system_zone(sb); |
d4c402d9 | 6477 | |
61bb4a1c TT |
6478 | if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb)) |
6479 | ext4_stop_mmpd(sbi); | |
6480 | ||
f25391eb LC |
6481 | /* |
6482 | * Some options can be enabled by ext4 and/or by VFS mount flag | |
6483 | * either way we need to make sure it matches in both *flags and | |
6484 | * s_flags. Copy those selected flags from s_flags to *flags | |
6485 | */ | |
6486 | *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags); | |
d4c402d9 | 6487 | |
ca9b404f RA |
6488 | ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s. Quota mode: %s.", |
6489 | orig_data, ext4_quota_mode(sb)); | |
d4c402d9 | 6490 | kfree(orig_data); |
ac27a0ec | 6491 | return 0; |
0b8e58a1 | 6492 | |
ac27a0ec DK |
6493 | restore_opts: |
6494 | sb->s_flags = old_sb_flags; | |
6495 | sbi->s_mount_opt = old_opts.s_mount_opt; | |
a2595b8a | 6496 | sbi->s_mount_opt2 = old_opts.s_mount_opt2; |
ac27a0ec DK |
6497 | sbi->s_resuid = old_opts.s_resuid; |
6498 | sbi->s_resgid = old_opts.s_resgid; | |
6499 | sbi->s_commit_interval = old_opts.s_commit_interval; | |
30773840 TT |
6500 | sbi->s_min_batch_time = old_opts.s_min_batch_time; |
6501 | sbi->s_max_batch_time = old_opts.s_max_batch_time; | |
dd0db94f | 6502 | if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) |
0f5bde1d | 6503 | ext4_release_system_zone(sb); |
ac27a0ec DK |
6504 | #ifdef CONFIG_QUOTA |
6505 | sbi->s_jquota_fmt = old_opts.s_jquota_fmt; | |
a2d4a646 | 6506 | for (i = 0; i < EXT4_MAXQUOTAS; i++) { |
33458eab TT |
6507 | to_free[i] = get_qf_name(sb, sbi, i); |
6508 | rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); | |
ac27a0ec | 6509 | } |
33458eab TT |
6510 | synchronize_rcu(); |
6511 | for (i = 0; i < EXT4_MAXQUOTAS; i++) | |
6512 | kfree(to_free[i]); | |
ac27a0ec | 6513 | #endif |
61bb4a1c TT |
6514 | if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb)) |
6515 | ext4_stop_mmpd(sbi); | |
d4c402d9 | 6516 | kfree(orig_data); |
ac27a0ec DK |
6517 | return err; |
6518 | } | |
6519 | ||
689c958c LX |
6520 | #ifdef CONFIG_QUOTA |
6521 | static int ext4_statfs_project(struct super_block *sb, | |
6522 | kprojid_t projid, struct kstatfs *buf) | |
6523 | { | |
6524 | struct kqid qid; | |
6525 | struct dquot *dquot; | |
6526 | u64 limit; | |
6527 | u64 curblock; | |
6528 | ||
6529 | qid = make_kqid_projid(projid); | |
6530 | dquot = dqget(sb, qid); | |
6531 | if (IS_ERR(dquot)) | |
6532 | return PTR_ERR(dquot); | |
7b9ca4c6 | 6533 | spin_lock(&dquot->dq_dqb_lock); |
689c958c | 6534 | |
a08fe66e CX |
6535 | limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, |
6536 | dquot->dq_dqb.dqb_bhardlimit); | |
57c32ea4 CX |
6537 | limit >>= sb->s_blocksize_bits; |
6538 | ||
689c958c | 6539 | if (limit && buf->f_blocks > limit) { |
f06925c7 KK |
6540 | curblock = (dquot->dq_dqb.dqb_curspace + |
6541 | dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; | |
689c958c LX |
6542 | buf->f_blocks = limit; |
6543 | buf->f_bfree = buf->f_bavail = | |
6544 | (buf->f_blocks > curblock) ? | |
6545 | (buf->f_blocks - curblock) : 0; | |
6546 | } | |
6547 | ||
a08fe66e CX |
6548 | limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, |
6549 | dquot->dq_dqb.dqb_ihardlimit); | |
689c958c LX |
6550 | if (limit && buf->f_files > limit) { |
6551 | buf->f_files = limit; | |
6552 | buf->f_ffree = | |
6553 | (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? | |
6554 | (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; | |
6555 | } | |
6556 | ||
7b9ca4c6 | 6557 | spin_unlock(&dquot->dq_dqb_lock); |
689c958c LX |
6558 | dqput(dquot); |
6559 | return 0; | |
6560 | } | |
6561 | #endif | |
6562 | ||
2b2d6d01 | 6563 | static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) |
ac27a0ec DK |
6564 | { |
6565 | struct super_block *sb = dentry->d_sb; | |
617ba13b MC |
6566 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
6567 | struct ext4_super_block *es = sbi->s_es; | |
27dd4385 | 6568 | ext4_fsblk_t overhead = 0, resv_blocks; |
d02a9391 | 6569 | s64 bfree; |
27dd4385 | 6570 | resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); |
ac27a0ec | 6571 | |
952fc18e TT |
6572 | if (!test_opt(sb, MINIX_DF)) |
6573 | overhead = sbi->s_overhead; | |
ac27a0ec | 6574 | |
617ba13b | 6575 | buf->f_type = EXT4_SUPER_MAGIC; |
ac27a0ec | 6576 | buf->f_bsize = sb->s_blocksize; |
b72f78cb | 6577 | buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); |
57042651 TT |
6578 | bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - |
6579 | percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); | |
d02a9391 | 6580 | /* prevent underflow in case that few free space is available */ |
57042651 | 6581 | buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); |
27dd4385 LC |
6582 | buf->f_bavail = buf->f_bfree - |
6583 | (ext4_r_blocks_count(es) + resv_blocks); | |
6584 | if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) | |
ac27a0ec DK |
6585 | buf->f_bavail = 0; |
6586 | buf->f_files = le32_to_cpu(es->s_inodes_count); | |
52d9f3b4 | 6587 | buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); |
617ba13b | 6588 | buf->f_namelen = EXT4_NAME_LEN; |
9591c3a3 | 6589 | buf->f_fsid = uuid_to_fsid(es->s_uuid); |
0b8e58a1 | 6590 | |
689c958c LX |
6591 | #ifdef CONFIG_QUOTA |
6592 | if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && | |
6593 | sb_has_quota_limits_enabled(sb, PRJQUOTA)) | |
6594 | ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); | |
6595 | #endif | |
ac27a0ec DK |
6596 | return 0; |
6597 | } | |
6598 | ||
ac27a0ec DK |
6599 | |
6600 | #ifdef CONFIG_QUOTA | |
6601 | ||
bc8230ee JK |
6602 | /* |
6603 | * Helper functions so that transaction is started before we acquire dqio_sem | |
6604 | * to keep correct lock ordering of transaction > dqio_sem | |
6605 | */ | |
ac27a0ec DK |
6606 | static inline struct inode *dquot_to_inode(struct dquot *dquot) |
6607 | { | |
4c376dca | 6608 | return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; |
ac27a0ec DK |
6609 | } |
6610 | ||
617ba13b | 6611 | static int ext4_write_dquot(struct dquot *dquot) |
ac27a0ec DK |
6612 | { |
6613 | int ret, err; | |
6614 | handle_t *handle; | |
6615 | struct inode *inode; | |
6616 | ||
6617 | inode = dquot_to_inode(dquot); | |
9924a92a | 6618 | handle = ext4_journal_start(inode, EXT4_HT_QUOTA, |
0b8e58a1 | 6619 | EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); |
ac27a0ec DK |
6620 | if (IS_ERR(handle)) |
6621 | return PTR_ERR(handle); | |
6622 | ret = dquot_commit(dquot); | |
617ba13b | 6623 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
6624 | if (!ret) |
6625 | ret = err; | |
6626 | return ret; | |
6627 | } | |
6628 | ||
617ba13b | 6629 | static int ext4_acquire_dquot(struct dquot *dquot) |
ac27a0ec DK |
6630 | { |
6631 | int ret, err; | |
6632 | handle_t *handle; | |
6633 | ||
9924a92a | 6634 | handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, |
0b8e58a1 | 6635 | EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); |
ac27a0ec DK |
6636 | if (IS_ERR(handle)) |
6637 | return PTR_ERR(handle); | |
6638 | ret = dquot_acquire(dquot); | |
617ba13b | 6639 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
6640 | if (!ret) |
6641 | ret = err; | |
6642 | return ret; | |
6643 | } | |
6644 | ||
617ba13b | 6645 | static int ext4_release_dquot(struct dquot *dquot) |
ac27a0ec DK |
6646 | { |
6647 | int ret, err; | |
6648 | handle_t *handle; | |
6649 | ||
9924a92a | 6650 | handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, |
0b8e58a1 | 6651 | EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); |
9c3013e9 JK |
6652 | if (IS_ERR(handle)) { |
6653 | /* Release dquot anyway to avoid endless cycle in dqput() */ | |
6654 | dquot_release(dquot); | |
ac27a0ec | 6655 | return PTR_ERR(handle); |
9c3013e9 | 6656 | } |
ac27a0ec | 6657 | ret = dquot_release(dquot); |
617ba13b | 6658 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
6659 | if (!ret) |
6660 | ret = err; | |
6661 | return ret; | |
6662 | } | |
6663 | ||
617ba13b | 6664 | static int ext4_mark_dquot_dirty(struct dquot *dquot) |
ac27a0ec | 6665 | { |
262b4662 | 6666 | struct super_block *sb = dquot->dq_sb; |
262b4662 | 6667 | |
f177ee08 | 6668 | if (ext4_is_quota_journalled(sb)) { |
ac27a0ec | 6669 | dquot_mark_dquot_dirty(dquot); |
617ba13b | 6670 | return ext4_write_dquot(dquot); |
ac27a0ec DK |
6671 | } else { |
6672 | return dquot_mark_dquot_dirty(dquot); | |
6673 | } | |
6674 | } | |
6675 | ||
617ba13b | 6676 | static int ext4_write_info(struct super_block *sb, int type) |
ac27a0ec DK |
6677 | { |
6678 | int ret, err; | |
6679 | handle_t *handle; | |
6680 | ||
6681 | /* Data block + inode block */ | |
2b0143b5 | 6682 | handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2); |
ac27a0ec DK |
6683 | if (IS_ERR(handle)) |
6684 | return PTR_ERR(handle); | |
6685 | ret = dquot_commit_info(sb, type); | |
617ba13b | 6686 | err = ext4_journal_stop(handle); |
ac27a0ec DK |
6687 | if (!ret) |
6688 | ret = err; | |
6689 | return ret; | |
6690 | } | |
6691 | ||
daf647d2 TT |
6692 | static void lockdep_set_quota_inode(struct inode *inode, int subclass) |
6693 | { | |
6694 | struct ext4_inode_info *ei = EXT4_I(inode); | |
6695 | ||
6696 | /* The first argument of lockdep_set_subclass has to be | |
6697 | * *exactly* the same as the argument to init_rwsem() --- in | |
6698 | * this case, in init_once() --- or lockdep gets unhappy | |
6699 | * because the name of the lock is set using the | |
6700 | * stringification of the argument to init_rwsem(). | |
6701 | */ | |
6702 | (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ | |
6703 | lockdep_set_subclass(&ei->i_data_sem, subclass); | |
6704 | } | |
6705 | ||
ac27a0ec DK |
6706 | /* |
6707 | * Standard function to be called on quota_on | |
6708 | */ | |
617ba13b | 6709 | static int ext4_quota_on(struct super_block *sb, int type, int format_id, |
8c54ca9c | 6710 | const struct path *path) |
ac27a0ec DK |
6711 | { |
6712 | int err; | |
ac27a0ec DK |
6713 | |
6714 | if (!test_opt(sb, QUOTA)) | |
6715 | return -EINVAL; | |
0623543b | 6716 | |
ac27a0ec | 6717 | /* Quotafile not on the same filesystem? */ |
d8c9584e | 6718 | if (path->dentry->d_sb != sb) |
ac27a0ec | 6719 | return -EXDEV; |
e0770e91 JK |
6720 | |
6721 | /* Quota already enabled for this file? */ | |
6722 | if (IS_NOQUOTA(d_inode(path->dentry))) | |
6723 | return -EBUSY; | |
6724 | ||
0623543b JK |
6725 | /* Journaling quota? */ |
6726 | if (EXT4_SB(sb)->s_qf_names[type]) { | |
2b2d6d01 | 6727 | /* Quotafile not in fs root? */ |
f00c9e44 | 6728 | if (path->dentry->d_parent != sb->s_root) |
b31e1552 ES |
6729 | ext4_msg(sb, KERN_WARNING, |
6730 | "Quota file not on filesystem root. " | |
6731 | "Journaled quota will not work"); | |
91389240 JK |
6732 | sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY; |
6733 | } else { | |
6734 | /* | |
6735 | * Clear the flag just in case mount options changed since | |
6736 | * last time. | |
6737 | */ | |
6738 | sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY; | |
2b2d6d01 | 6739 | } |
0623543b JK |
6740 | |
6741 | /* | |
6742 | * When we journal data on quota file, we have to flush journal to see | |
6743 | * all updates to the file when we bypass pagecache... | |
6744 | */ | |
0390131b | 6745 | if (EXT4_SB(sb)->s_journal && |
2b0143b5 | 6746 | ext4_should_journal_data(d_inode(path->dentry))) { |
0623543b JK |
6747 | /* |
6748 | * We don't need to lock updates but journal_flush() could | |
6749 | * otherwise be livelocked... | |
6750 | */ | |
6751 | jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); | |
01d5d965 | 6752 | err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0); |
0623543b | 6753 | jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); |
f00c9e44 | 6754 | if (err) |
7ffe1ea8 | 6755 | return err; |
0623543b | 6756 | } |
957153fc | 6757 | |
daf647d2 TT |
6758 | lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); |
6759 | err = dquot_quota_on(sb, type, format_id, path); | |
957153fc | 6760 | if (err) { |
daf647d2 TT |
6761 | lockdep_set_quota_inode(path->dentry->d_inode, |
6762 | I_DATA_SEM_NORMAL); | |
957153fc JK |
6763 | } else { |
6764 | struct inode *inode = d_inode(path->dentry); | |
6765 | handle_t *handle; | |
6766 | ||
61a92987 JK |
6767 | /* |
6768 | * Set inode flags to prevent userspace from messing with quota | |
6769 | * files. If this fails, we return success anyway since quotas | |
6770 | * are already enabled and this is not a hard failure. | |
6771 | */ | |
957153fc JK |
6772 | inode_lock(inode); |
6773 | handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); | |
6774 | if (IS_ERR(handle)) | |
6775 | goto unlock_inode; | |
6776 | EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL; | |
6777 | inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, | |
6778 | S_NOATIME | S_IMMUTABLE); | |
4209ae12 | 6779 | err = ext4_mark_inode_dirty(handle, inode); |
957153fc JK |
6780 | ext4_journal_stop(handle); |
6781 | unlock_inode: | |
6782 | inode_unlock(inode); | |
6783 | } | |
daf647d2 | 6784 | return err; |
ac27a0ec DK |
6785 | } |
6786 | ||
7c319d32 AK |
6787 | static int ext4_quota_enable(struct super_block *sb, int type, int format_id, |
6788 | unsigned int flags) | |
6789 | { | |
6790 | int err; | |
6791 | struct inode *qf_inode; | |
a2d4a646 | 6792 | unsigned long qf_inums[EXT4_MAXQUOTAS] = { |
7c319d32 | 6793 | le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), |
689c958c LX |
6794 | le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), |
6795 | le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) | |
7c319d32 AK |
6796 | }; |
6797 | ||
e2b911c5 | 6798 | BUG_ON(!ext4_has_feature_quota(sb)); |
7c319d32 AK |
6799 | |
6800 | if (!qf_inums[type]) | |
6801 | return -EPERM; | |
6802 | ||
8a363970 | 6803 | qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL); |
7c319d32 AK |
6804 | if (IS_ERR(qf_inode)) { |
6805 | ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); | |
6806 | return PTR_ERR(qf_inode); | |
6807 | } | |
6808 | ||
bcb13850 JK |
6809 | /* Don't account quota for quota files to avoid recursion */ |
6810 | qf_inode->i_flags |= S_NOQUOTA; | |
daf647d2 | 6811 | lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); |
7212b95e | 6812 | err = dquot_load_quota_inode(qf_inode, type, format_id, flags); |
daf647d2 TT |
6813 | if (err) |
6814 | lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); | |
61157b24 | 6815 | iput(qf_inode); |
7c319d32 AK |
6816 | |
6817 | return err; | |
6818 | } | |
6819 | ||
6820 | /* Enable usage tracking for all quota types. */ | |
25c6d98f | 6821 | int ext4_enable_quotas(struct super_block *sb) |
7c319d32 AK |
6822 | { |
6823 | int type, err = 0; | |
a2d4a646 | 6824 | unsigned long qf_inums[EXT4_MAXQUOTAS] = { |
7c319d32 | 6825 | le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), |
689c958c LX |
6826 | le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), |
6827 | le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) | |
7c319d32 | 6828 | }; |
49da9392 JK |
6829 | bool quota_mopt[EXT4_MAXQUOTAS] = { |
6830 | test_opt(sb, USRQUOTA), | |
6831 | test_opt(sb, GRPQUOTA), | |
6832 | test_opt(sb, PRJQUOTA), | |
6833 | }; | |
7c319d32 | 6834 | |
91389240 | 6835 | sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY; |
a2d4a646 | 6836 | for (type = 0; type < EXT4_MAXQUOTAS; type++) { |
7c319d32 AK |
6837 | if (qf_inums[type]) { |
6838 | err = ext4_quota_enable(sb, type, QFMT_VFS_V1, | |
49da9392 JK |
6839 | DQUOT_USAGE_ENABLED | |
6840 | (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); | |
7c319d32 AK |
6841 | if (err) { |
6842 | ext4_warning(sb, | |
72ba7450 TT |
6843 | "Failed to enable quota tracking " |
6844 | "(type=%d, err=%d). Please run " | |
6845 | "e2fsck to fix.", type, err); | |
7f144fd0 JU |
6846 | for (type--; type >= 0; type--) |
6847 | dquot_quota_off(sb, type); | |
6848 | ||
7c319d32 AK |
6849 | return err; |
6850 | } | |
6851 | } | |
6852 | } | |
6853 | return 0; | |
6854 | } | |
6855 | ||
ca0e05e4 DM |
6856 | static int ext4_quota_off(struct super_block *sb, int type) |
6857 | { | |
21f97697 JK |
6858 | struct inode *inode = sb_dqopt(sb)->files[type]; |
6859 | handle_t *handle; | |
957153fc | 6860 | int err; |
21f97697 | 6861 | |
87009d86 DM |
6862 | /* Force all delayed allocation blocks to be allocated. |
6863 | * Caller already holds s_umount sem */ | |
6864 | if (test_opt(sb, DELALLOC)) | |
ca0e05e4 | 6865 | sync_filesystem(sb); |
ca0e05e4 | 6866 | |
957153fc | 6867 | if (!inode || !igrab(inode)) |
0b268590 AG |
6868 | goto out; |
6869 | ||
957153fc | 6870 | err = dquot_quota_off(sb, type); |
964edf66 | 6871 | if (err || ext4_has_feature_quota(sb)) |
957153fc JK |
6872 | goto out_put; |
6873 | ||
6874 | inode_lock(inode); | |
61a92987 JK |
6875 | /* |
6876 | * Update modification times of quota files when userspace can | |
6877 | * start looking at them. If we fail, we return success anyway since | |
6878 | * this is not a hard failure and quotas are already disabled. | |
6879 | */ | |
9924a92a | 6880 | handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); |
4209ae12 HS |
6881 | if (IS_ERR(handle)) { |
6882 | err = PTR_ERR(handle); | |
957153fc | 6883 | goto out_unlock; |
4209ae12 | 6884 | } |
957153fc JK |
6885 | EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL); |
6886 | inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); | |
eeca7ea1 | 6887 | inode->i_mtime = inode->i_ctime = current_time(inode); |
4209ae12 | 6888 | err = ext4_mark_inode_dirty(handle, inode); |
21f97697 | 6889 | ext4_journal_stop(handle); |
957153fc JK |
6890 | out_unlock: |
6891 | inode_unlock(inode); | |
6892 | out_put: | |
964edf66 | 6893 | lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL); |
957153fc JK |
6894 | iput(inode); |
6895 | return err; | |
21f97697 | 6896 | out: |
ca0e05e4 DM |
6897 | return dquot_quota_off(sb, type); |
6898 | } | |
6899 | ||
ac27a0ec DK |
6900 | /* Read data from quotafile - avoid pagecache and such because we cannot afford |
6901 | * acquiring the locks... As quota files are never truncated and quota code | |
25985edc | 6902 | * itself serializes the operations (and no one else should touch the files) |
ac27a0ec | 6903 | * we don't have to be afraid of races */ |
617ba13b | 6904 | static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, |
ac27a0ec DK |
6905 | size_t len, loff_t off) |
6906 | { | |
6907 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
725d26d3 | 6908 | ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); |
ac27a0ec DK |
6909 | int offset = off & (sb->s_blocksize - 1); |
6910 | int tocopy; | |
6911 | size_t toread; | |
6912 | struct buffer_head *bh; | |
6913 | loff_t i_size = i_size_read(inode); | |
6914 | ||
6915 | if (off > i_size) | |
6916 | return 0; | |
6917 | if (off+len > i_size) | |
6918 | len = i_size-off; | |
6919 | toread = len; | |
6920 | while (toread > 0) { | |
6921 | tocopy = sb->s_blocksize - offset < toread ? | |
6922 | sb->s_blocksize - offset : toread; | |
1c215028 TT |
6923 | bh = ext4_bread(NULL, inode, blk, 0); |
6924 | if (IS_ERR(bh)) | |
6925 | return PTR_ERR(bh); | |
ac27a0ec DK |
6926 | if (!bh) /* A hole? */ |
6927 | memset(data, 0, tocopy); | |
6928 | else | |
6929 | memcpy(data, bh->b_data+offset, tocopy); | |
6930 | brelse(bh); | |
6931 | offset = 0; | |
6932 | toread -= tocopy; | |
6933 | data += tocopy; | |
6934 | blk++; | |
6935 | } | |
6936 | return len; | |
6937 | } | |
6938 | ||
6939 | /* Write to quotafile (we know the transaction is already started and has | |
6940 | * enough credits) */ | |
617ba13b | 6941 | static ssize_t ext4_quota_write(struct super_block *sb, int type, |
ac27a0ec DK |
6942 | const char *data, size_t len, loff_t off) |
6943 | { | |
6944 | struct inode *inode = sb_dqopt(sb)->files[type]; | |
725d26d3 | 6945 | ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); |
4209ae12 | 6946 | int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1); |
c5e298ae | 6947 | int retries = 0; |
ac27a0ec DK |
6948 | struct buffer_head *bh; |
6949 | handle_t *handle = journal_current_handle(); | |
6950 | ||
0390131b | 6951 | if (EXT4_SB(sb)->s_journal && !handle) { |
b31e1552 ES |
6952 | ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" |
6953 | " cancelled because transaction is not started", | |
9c3013e9 JK |
6954 | (unsigned long long)off, (unsigned long long)len); |
6955 | return -EIO; | |
6956 | } | |
67eeb568 DM |
6957 | /* |
6958 | * Since we account only one data block in transaction credits, | |
6959 | * then it is impossible to cross a block boundary. | |
6960 | */ | |
6961 | if (sb->s_blocksize - offset < len) { | |
6962 | ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" | |
6963 | " cancelled because not block aligned", | |
6964 | (unsigned long long)off, (unsigned long long)len); | |
6965 | return -EIO; | |
6966 | } | |
6967 | ||
c5e298ae TT |
6968 | do { |
6969 | bh = ext4_bread(handle, inode, blk, | |
6970 | EXT4_GET_BLOCKS_CREATE | | |
6971 | EXT4_GET_BLOCKS_METADATA_NOFAIL); | |
45586c70 | 6972 | } while (PTR_ERR(bh) == -ENOSPC && |
c5e298ae | 6973 | ext4_should_retry_alloc(inode->i_sb, &retries)); |
1c215028 TT |
6974 | if (IS_ERR(bh)) |
6975 | return PTR_ERR(bh); | |
67eeb568 DM |
6976 | if (!bh) |
6977 | goto out; | |
5d601255 | 6978 | BUFFER_TRACE(bh, "get write access"); |
188c299e | 6979 | err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE); |
62d2b5f2 JK |
6980 | if (err) { |
6981 | brelse(bh); | |
1c215028 | 6982 | return err; |
ac27a0ec | 6983 | } |
67eeb568 DM |
6984 | lock_buffer(bh); |
6985 | memcpy(bh->b_data+offset, data, len); | |
6986 | flush_dcache_page(bh->b_page); | |
6987 | unlock_buffer(bh); | |
62d2b5f2 | 6988 | err = ext4_handle_dirty_metadata(handle, NULL, bh); |
67eeb568 | 6989 | brelse(bh); |
ac27a0ec | 6990 | out: |
67eeb568 DM |
6991 | if (inode->i_size < off + len) { |
6992 | i_size_write(inode, off + len); | |
617ba13b | 6993 | EXT4_I(inode)->i_disksize = inode->i_size; |
4209ae12 HS |
6994 | err2 = ext4_mark_inode_dirty(handle, inode); |
6995 | if (unlikely(err2 && !err)) | |
6996 | err = err2; | |
ac27a0ec | 6997 | } |
4209ae12 | 6998 | return err ? err : len; |
ac27a0ec | 6999 | } |
ac27a0ec DK |
7000 | #endif |
7001 | ||
152a0836 AV |
7002 | static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, |
7003 | const char *dev_name, void *data) | |
ac27a0ec | 7004 | { |
152a0836 | 7005 | return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); |
ac27a0ec DK |
7006 | } |
7007 | ||
c290ea01 | 7008 | #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) |
24b58424 TT |
7009 | static inline void register_as_ext2(void) |
7010 | { | |
7011 | int err = register_filesystem(&ext2_fs_type); | |
7012 | if (err) | |
7013 | printk(KERN_WARNING | |
7014 | "EXT4-fs: Unable to register as ext2 (%d)\n", err); | |
7015 | } | |
7016 | ||
7017 | static inline void unregister_as_ext2(void) | |
7018 | { | |
7019 | unregister_filesystem(&ext2_fs_type); | |
7020 | } | |
2035e776 TT |
7021 | |
7022 | static inline int ext2_feature_set_ok(struct super_block *sb) | |
7023 | { | |
e2b911c5 | 7024 | if (ext4_has_unknown_ext2_incompat_features(sb)) |
2035e776 | 7025 | return 0; |
bc98a42c | 7026 | if (sb_rdonly(sb)) |
2035e776 | 7027 | return 1; |
e2b911c5 | 7028 | if (ext4_has_unknown_ext2_ro_compat_features(sb)) |
2035e776 TT |
7029 | return 0; |
7030 | return 1; | |
7031 | } | |
24b58424 TT |
7032 | #else |
7033 | static inline void register_as_ext2(void) { } | |
7034 | static inline void unregister_as_ext2(void) { } | |
2035e776 | 7035 | static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } |
24b58424 TT |
7036 | #endif |
7037 | ||
24b58424 TT |
7038 | static inline void register_as_ext3(void) |
7039 | { | |
7040 | int err = register_filesystem(&ext3_fs_type); | |
7041 | if (err) | |
7042 | printk(KERN_WARNING | |
7043 | "EXT4-fs: Unable to register as ext3 (%d)\n", err); | |
7044 | } | |
7045 | ||
7046 | static inline void unregister_as_ext3(void) | |
7047 | { | |
7048 | unregister_filesystem(&ext3_fs_type); | |
7049 | } | |
2035e776 TT |
7050 | |
7051 | static inline int ext3_feature_set_ok(struct super_block *sb) | |
7052 | { | |
e2b911c5 | 7053 | if (ext4_has_unknown_ext3_incompat_features(sb)) |
2035e776 | 7054 | return 0; |
e2b911c5 | 7055 | if (!ext4_has_feature_journal(sb)) |
2035e776 | 7056 | return 0; |
bc98a42c | 7057 | if (sb_rdonly(sb)) |
2035e776 | 7058 | return 1; |
e2b911c5 | 7059 | if (ext4_has_unknown_ext3_ro_compat_features(sb)) |
2035e776 TT |
7060 | return 0; |
7061 | return 1; | |
7062 | } | |
24b58424 | 7063 | |
03010a33 TT |
7064 | static struct file_system_type ext4_fs_type = { |
7065 | .owner = THIS_MODULE, | |
7066 | .name = "ext4", | |
152a0836 | 7067 | .mount = ext4_mount, |
03010a33 | 7068 | .kill_sb = kill_block_super, |
14f3db55 | 7069 | .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP, |
03010a33 | 7070 | }; |
7f78e035 | 7071 | MODULE_ALIAS_FS("ext4"); |
03010a33 | 7072 | |
e9e3bcec ES |
7073 | /* Shared across all ext4 file systems */ |
7074 | wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; | |
e9e3bcec | 7075 | |
5dabfc78 | 7076 | static int __init ext4_init_fs(void) |
ac27a0ec | 7077 | { |
e9e3bcec | 7078 | int i, err; |
c9de560d | 7079 | |
e294a537 | 7080 | ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); |
07c0c5d8 | 7081 | ext4_li_info = NULL; |
07c0c5d8 | 7082 | |
9a4c8019 | 7083 | /* Build-time check for flags consistency */ |
12e9b892 | 7084 | ext4_check_flag_values(); |
e9e3bcec | 7085 | |
e142d052 | 7086 | for (i = 0; i < EXT4_WQ_HASH_SZ; i++) |
e9e3bcec | 7087 | init_waitqueue_head(&ext4__ioend_wq[i]); |
e9e3bcec | 7088 | |
51865fda | 7089 | err = ext4_init_es(); |
6fd058f7 TT |
7090 | if (err) |
7091 | return err; | |
51865fda | 7092 | |
1dc0aa46 | 7093 | err = ext4_init_pending(); |
22cfe4b4 EB |
7094 | if (err) |
7095 | goto out7; | |
7096 | ||
7097 | err = ext4_init_post_read_processing(); | |
1dc0aa46 EW |
7098 | if (err) |
7099 | goto out6; | |
7100 | ||
51865fda ZL |
7101 | err = ext4_init_pageio(); |
7102 | if (err) | |
b5799018 | 7103 | goto out5; |
51865fda | 7104 | |
5dabfc78 | 7105 | err = ext4_init_system_zone(); |
bd2d0210 | 7106 | if (err) |
b5799018 | 7107 | goto out4; |
857ac889 | 7108 | |
b5799018 | 7109 | err = ext4_init_sysfs(); |
dd68314c | 7110 | if (err) |
b5799018 | 7111 | goto out3; |
857ac889 | 7112 | |
5dabfc78 | 7113 | err = ext4_init_mballoc(); |
c9de560d AT |
7114 | if (err) |
7115 | goto out2; | |
ac27a0ec DK |
7116 | err = init_inodecache(); |
7117 | if (err) | |
7118 | goto out1; | |
aa75f4d3 HS |
7119 | |
7120 | err = ext4_fc_init_dentry_cache(); | |
7121 | if (err) | |
7122 | goto out05; | |
7123 | ||
24b58424 | 7124 | register_as_ext3(); |
2035e776 | 7125 | register_as_ext2(); |
03010a33 | 7126 | err = register_filesystem(&ext4_fs_type); |
ac27a0ec DK |
7127 | if (err) |
7128 | goto out; | |
bfff6873 | 7129 | |
ac27a0ec DK |
7130 | return 0; |
7131 | out: | |
24b58424 TT |
7132 | unregister_as_ext2(); |
7133 | unregister_as_ext3(); | |
aa75f4d3 | 7134 | out05: |
ac27a0ec DK |
7135 | destroy_inodecache(); |
7136 | out1: | |
5dabfc78 | 7137 | ext4_exit_mballoc(); |
9c191f70 | 7138 | out2: |
b5799018 TT |
7139 | ext4_exit_sysfs(); |
7140 | out3: | |
5dabfc78 | 7141 | ext4_exit_system_zone(); |
b5799018 | 7142 | out4: |
5dabfc78 | 7143 | ext4_exit_pageio(); |
b5799018 | 7144 | out5: |
22cfe4b4 | 7145 | ext4_exit_post_read_processing(); |
1dc0aa46 | 7146 | out6: |
22cfe4b4 EB |
7147 | ext4_exit_pending(); |
7148 | out7: | |
51865fda ZL |
7149 | ext4_exit_es(); |
7150 | ||
ac27a0ec DK |
7151 | return err; |
7152 | } | |
7153 | ||
5dabfc78 | 7154 | static void __exit ext4_exit_fs(void) |
ac27a0ec | 7155 | { |
bfff6873 | 7156 | ext4_destroy_lazyinit_thread(); |
24b58424 TT |
7157 | unregister_as_ext2(); |
7158 | unregister_as_ext3(); | |
03010a33 | 7159 | unregister_filesystem(&ext4_fs_type); |
ac27a0ec | 7160 | destroy_inodecache(); |
5dabfc78 | 7161 | ext4_exit_mballoc(); |
b5799018 | 7162 | ext4_exit_sysfs(); |
5dabfc78 TT |
7163 | ext4_exit_system_zone(); |
7164 | ext4_exit_pageio(); | |
22cfe4b4 | 7165 | ext4_exit_post_read_processing(); |
dd12ed14 | 7166 | ext4_exit_es(); |
1dc0aa46 | 7167 | ext4_exit_pending(); |
ac27a0ec DK |
7168 | } |
7169 | ||
7170 | MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); | |
83982b6f | 7171 | MODULE_DESCRIPTION("Fourth Extended Filesystem"); |
ac27a0ec | 7172 | MODULE_LICENSE("GPL"); |
7ef79ad5 | 7173 | MODULE_SOFTDEP("pre: crc32c"); |
5dabfc78 TT |
7174 | module_init(ext4_init_fs) |
7175 | module_exit(ext4_exit_fs) |