Commit | Line | Data |
---|---|---|
783f6184 RK |
1 | /* |
2 | * super.c - NILFS module and super block management. | |
3 | * | |
4 | * Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation. | |
5 | * | |
6 | * This program is free software; you can redistribute it and/or modify | |
7 | * it under the terms of the GNU General Public License as published by | |
8 | * the Free Software Foundation; either version 2 of the License, or | |
9 | * (at your option) any later version. | |
10 | * | |
11 | * This program is distributed in the hope that it will be useful, | |
12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
14 | * GNU General Public License for more details. | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License | |
17 | * along with this program; if not, write to the Free Software | |
18 | * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
19 | * | |
20 | * Written by Ryusuke Konishi <ryusuke@osrg.net> | |
21 | */ | |
22 | /* | |
23 | * linux/fs/ext2/super.c | |
24 | * | |
25 | * Copyright (C) 1992, 1993, 1994, 1995 | |
26 | * Remy Card (card@masi.ibp.fr) | |
27 | * Laboratoire MASI - Institut Blaise Pascal | |
28 | * Universite Pierre et Marie Curie (Paris VI) | |
29 | * | |
30 | * from | |
31 | * | |
32 | * linux/fs/minix/inode.c | |
33 | * | |
34 | * Copyright (C) 1991, 1992 Linus Torvalds | |
35 | * | |
36 | * Big-endian to little-endian byte-swapping/bitmaps by | |
37 | * David S. Miller (davem@caip.rutgers.edu), 1995 | |
38 | */ | |
39 | ||
40 | #include <linux/module.h> | |
41 | #include <linux/string.h> | |
42 | #include <linux/slab.h> | |
43 | #include <linux/init.h> | |
44 | #include <linux/blkdev.h> | |
45 | #include <linux/parser.h> | |
46 | #include <linux/random.h> | |
47 | #include <linux/crc32.h> | |
48 | #include <linux/smp_lock.h> | |
49 | #include <linux/vfs.h> | |
50 | #include <linux/writeback.h> | |
51 | #include <linux/kobject.h> | |
52 | #include <linux/exportfs.h> | |
b58a285b JS |
53 | #include <linux/seq_file.h> |
54 | #include <linux/mount.h> | |
783f6184 RK |
55 | #include "nilfs.h" |
56 | #include "mdt.h" | |
57 | #include "alloc.h" | |
05d0e94b RK |
58 | #include "btree.h" |
59 | #include "btnode.h" | |
783f6184 RK |
60 | #include "page.h" |
61 | #include "cpfile.h" | |
62 | #include "ifile.h" | |
63 | #include "dat.h" | |
64 | #include "segment.h" | |
65 | #include "segbuf.h" | |
66 | ||
67 | MODULE_AUTHOR("NTT Corp."); | |
68 | MODULE_DESCRIPTION("A New Implementation of the Log-structured Filesystem " | |
69 | "(NILFS)"); | |
783f6184 RK |
70 | MODULE_LICENSE("GPL"); |
71 | ||
41c88bd7 LH |
72 | struct kmem_cache *nilfs_inode_cachep; |
73 | struct kmem_cache *nilfs_transaction_cachep; | |
74 | struct kmem_cache *nilfs_segbuf_cachep; | |
75 | struct kmem_cache *nilfs_btree_path_cache; | |
76 | ||
783f6184 | 77 | static int nilfs_remount(struct super_block *sb, int *flags, char *data); |
783f6184 | 78 | |
c8a11c8a RK |
79 | static void nilfs_set_error(struct nilfs_sb_info *sbi) |
80 | { | |
81 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
d26493b6 | 82 | struct nilfs_super_block **sbp; |
c8a11c8a RK |
83 | |
84 | down_write(&nilfs->ns_sem); | |
85 | if (!(nilfs->ns_mount_state & NILFS_ERROR_FS)) { | |
86 | nilfs->ns_mount_state |= NILFS_ERROR_FS; | |
b2ac86e1 | 87 | sbp = nilfs_prepare_super(sbi, 0); |
d26493b6 JS |
88 | if (likely(sbp)) { |
89 | sbp[0]->s_state |= cpu_to_le16(NILFS_ERROR_FS); | |
b2ac86e1 JS |
90 | if (sbp[1]) |
91 | sbp[1]->s_state |= cpu_to_le16(NILFS_ERROR_FS); | |
92 | nilfs_commit_super(sbi, NILFS_SB_COMMIT_ALL); | |
d26493b6 | 93 | } |
c8a11c8a RK |
94 | } |
95 | up_write(&nilfs->ns_sem); | |
96 | } | |
97 | ||
783f6184 RK |
98 | /** |
99 | * nilfs_error() - report failure condition on a filesystem | |
100 | * | |
101 | * nilfs_error() sets an ERROR_FS flag on the superblock as well as | |
102 | * reporting an error message. It should be called when NILFS detects | |
103 | * incoherences or defects of meta data on disk. As for sustainable | |
104 | * errors such as a single-shot I/O error, nilfs_warning() or the printk() | |
105 | * function should be used instead. | |
106 | * | |
107 | * The segment constructor must not call this function because it can | |
108 | * kill itself. | |
109 | */ | |
110 | void nilfs_error(struct super_block *sb, const char *function, | |
111 | const char *fmt, ...) | |
112 | { | |
113 | struct nilfs_sb_info *sbi = NILFS_SB(sb); | |
114 | va_list args; | |
115 | ||
116 | va_start(args, fmt); | |
117 | printk(KERN_CRIT "NILFS error (device %s): %s: ", sb->s_id, function); | |
118 | vprintk(fmt, args); | |
119 | printk("\n"); | |
120 | va_end(args); | |
121 | ||
122 | if (!(sb->s_flags & MS_RDONLY)) { | |
c8a11c8a | 123 | nilfs_set_error(sbi); |
783f6184 RK |
124 | |
125 | if (nilfs_test_opt(sbi, ERRORS_RO)) { | |
126 | printk(KERN_CRIT "Remounting filesystem read-only\n"); | |
127 | sb->s_flags |= MS_RDONLY; | |
128 | } | |
129 | } | |
130 | ||
131 | if (nilfs_test_opt(sbi, ERRORS_PANIC)) | |
132 | panic("NILFS (device %s): panic forced after error\n", | |
133 | sb->s_id); | |
134 | } | |
135 | ||
136 | void nilfs_warning(struct super_block *sb, const char *function, | |
137 | const char *fmt, ...) | |
138 | { | |
139 | va_list args; | |
140 | ||
141 | va_start(args, fmt); | |
142 | printk(KERN_WARNING "NILFS warning (device %s): %s: ", | |
143 | sb->s_id, function); | |
144 | vprintk(fmt, args); | |
145 | printk("\n"); | |
146 | va_end(args); | |
147 | } | |
148 | ||
783f6184 | 149 | |
a53b4751 | 150 | struct inode *nilfs_alloc_inode_common(struct the_nilfs *nilfs) |
783f6184 RK |
151 | { |
152 | struct nilfs_inode_info *ii; | |
153 | ||
154 | ii = kmem_cache_alloc(nilfs_inode_cachep, GFP_NOFS); | |
155 | if (!ii) | |
156 | return NULL; | |
157 | ii->i_bh = NULL; | |
158 | ii->i_state = 0; | |
159 | ii->vfs_inode.i_version = 1; | |
a53b4751 | 160 | nilfs_btnode_cache_init(&ii->i_btnode_cache, nilfs->ns_bdi); |
783f6184 RK |
161 | return &ii->vfs_inode; |
162 | } | |
163 | ||
a53b4751 RK |
164 | struct inode *nilfs_alloc_inode(struct super_block *sb) |
165 | { | |
166 | return nilfs_alloc_inode_common(NILFS_SB(sb)->s_nilfs); | |
167 | } | |
168 | ||
783f6184 RK |
169 | void nilfs_destroy_inode(struct inode *inode) |
170 | { | |
171 | kmem_cache_free(nilfs_inode_cachep, NILFS_I(inode)); | |
172 | } | |
173 | ||
b2ac86e1 | 174 | static int nilfs_sync_super(struct nilfs_sb_info *sbi, int flag) |
783f6184 RK |
175 | { |
176 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
177 | int err; | |
178 | int barrier_done = 0; | |
179 | ||
180 | if (nilfs_test_opt(sbi, BARRIER)) { | |
e339ad31 | 181 | set_buffer_ordered(nilfs->ns_sbh[0]); |
783f6184 RK |
182 | barrier_done = 1; |
183 | } | |
184 | retry: | |
e339ad31 RK |
185 | set_buffer_dirty(nilfs->ns_sbh[0]); |
186 | err = sync_dirty_buffer(nilfs->ns_sbh[0]); | |
783f6184 RK |
187 | if (err == -EOPNOTSUPP && barrier_done) { |
188 | nilfs_warning(sbi->s_super, __func__, | |
189 | "barrier-based sync failed. " | |
190 | "disabling barriers\n"); | |
191 | nilfs_clear_opt(sbi, BARRIER); | |
192 | barrier_done = 0; | |
e339ad31 | 193 | clear_buffer_ordered(nilfs->ns_sbh[0]); |
783f6184 RK |
194 | goto retry; |
195 | } | |
e339ad31 | 196 | if (unlikely(err)) { |
783f6184 RK |
197 | printk(KERN_ERR |
198 | "NILFS: unable to write superblock (err=%d)\n", err); | |
e339ad31 | 199 | if (err == -EIO && nilfs->ns_sbh[1]) { |
b2ac86e1 JS |
200 | /* |
201 | * sbp[0] points to newer log than sbp[1], | |
202 | * so copy sbp[0] to sbp[1] to take over sbp[0]. | |
203 | */ | |
204 | memcpy(nilfs->ns_sbp[1], nilfs->ns_sbp[0], | |
205 | nilfs->ns_sbsize); | |
e339ad31 RK |
206 | nilfs_fall_back_super_block(nilfs); |
207 | goto retry; | |
208 | } | |
209 | } else { | |
210 | struct nilfs_super_block *sbp = nilfs->ns_sbp[0]; | |
211 | ||
b2ac86e1 JS |
212 | nilfs->ns_sbwcount++; |
213 | ||
e339ad31 RK |
214 | /* |
215 | * The latest segment becomes trailable from the position | |
216 | * written in superblock. | |
217 | */ | |
783f6184 | 218 | clear_nilfs_discontinued(nilfs); |
e339ad31 RK |
219 | |
220 | /* update GC protection for recent segments */ | |
221 | if (nilfs->ns_sbh[1]) { | |
b2ac86e1 | 222 | if (flag == NILFS_SB_COMMIT_ALL) { |
e339ad31 | 223 | set_buffer_dirty(nilfs->ns_sbh[1]); |
b2ac86e1 JS |
224 | if (sync_dirty_buffer(nilfs->ns_sbh[1]) < 0) |
225 | goto out; | |
e339ad31 | 226 | } |
b2ac86e1 JS |
227 | if (le64_to_cpu(nilfs->ns_sbp[1]->s_last_cno) < |
228 | le64_to_cpu(nilfs->ns_sbp[0]->s_last_cno)) | |
229 | sbp = nilfs->ns_sbp[1]; | |
e339ad31 | 230 | } |
783f6184 | 231 | |
b2ac86e1 JS |
232 | spin_lock(&nilfs->ns_last_segment_lock); |
233 | nilfs->ns_prot_seq = le64_to_cpu(sbp->s_last_seq); | |
234 | spin_unlock(&nilfs->ns_last_segment_lock); | |
235 | } | |
236 | out: | |
783f6184 RK |
237 | return err; |
238 | } | |
239 | ||
60f46b7e RK |
240 | void nilfs_set_log_cursor(struct nilfs_super_block *sbp, |
241 | struct the_nilfs *nilfs) | |
242 | { | |
243 | sector_t nfreeblocks; | |
244 | ||
245 | /* nilfs->ns_sem must be locked by the caller. */ | |
246 | nilfs_count_free_blocks(nilfs, &nfreeblocks); | |
247 | sbp->s_free_blocks_count = cpu_to_le64(nfreeblocks); | |
248 | ||
249 | spin_lock(&nilfs->ns_last_segment_lock); | |
250 | sbp->s_last_seq = cpu_to_le64(nilfs->ns_last_seq); | |
251 | sbp->s_last_pseg = cpu_to_le64(nilfs->ns_last_pseg); | |
252 | sbp->s_last_cno = cpu_to_le64(nilfs->ns_last_cno); | |
253 | spin_unlock(&nilfs->ns_last_segment_lock); | |
254 | } | |
255 | ||
b2ac86e1 JS |
256 | struct nilfs_super_block **nilfs_prepare_super(struct nilfs_sb_info *sbi, |
257 | int flip) | |
783f6184 RK |
258 | { |
259 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
e339ad31 | 260 | struct nilfs_super_block **sbp = nilfs->ns_sbp; |
783f6184 | 261 | |
d26493b6 | 262 | /* nilfs->ns_sem must be locked by the caller. */ |
34cb9b5c | 263 | if (sbp[0]->s_magic != cpu_to_le16(NILFS_SUPER_MAGIC)) { |
d26493b6 JS |
264 | if (sbp[1] && |
265 | sbp[1]->s_magic == cpu_to_le16(NILFS_SUPER_MAGIC)) { | |
b2ac86e1 | 266 | memcpy(sbp[0], sbp[1], nilfs->ns_sbsize); |
d26493b6 | 267 | } else { |
e339ad31 RK |
268 | printk(KERN_CRIT "NILFS: superblock broke on dev %s\n", |
269 | sbi->s_super->s_id); | |
d26493b6 | 270 | return NULL; |
e339ad31 | 271 | } |
b2ac86e1 JS |
272 | } else if (sbp[1] && |
273 | sbp[1]->s_magic != cpu_to_le16(NILFS_SUPER_MAGIC)) { | |
274 | memcpy(sbp[1], sbp[0], nilfs->ns_sbsize); | |
e339ad31 | 275 | } |
b2ac86e1 JS |
276 | |
277 | if (flip && sbp[1]) | |
278 | nilfs_swap_super_block(nilfs); | |
279 | ||
d26493b6 JS |
280 | return sbp; |
281 | } | |
282 | ||
b2ac86e1 | 283 | int nilfs_commit_super(struct nilfs_sb_info *sbi, int flag) |
d26493b6 JS |
284 | { |
285 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
286 | struct nilfs_super_block **sbp = nilfs->ns_sbp; | |
287 | time_t t; | |
288 | ||
289 | /* nilfs->ns_sem must be locked by the caller. */ | |
e339ad31 | 290 | t = get_seconds(); |
b2ac86e1 | 291 | nilfs->ns_sbwtime = t; |
e339ad31 RK |
292 | sbp[0]->s_wtime = cpu_to_le64(t); |
293 | sbp[0]->s_sum = 0; | |
294 | sbp[0]->s_sum = cpu_to_le32(crc32_le(nilfs->ns_crc_seed, | |
295 | (unsigned char *)sbp[0], | |
296 | nilfs->ns_sbsize)); | |
b2ac86e1 JS |
297 | if (flag == NILFS_SB_COMMIT_ALL && sbp[1]) { |
298 | sbp[1]->s_wtime = sbp[0]->s_wtime; | |
299 | sbp[1]->s_sum = 0; | |
300 | sbp[1]->s_sum = cpu_to_le32(crc32_le(nilfs->ns_crc_seed, | |
301 | (unsigned char *)sbp[1], | |
302 | nilfs->ns_sbsize)); | |
e339ad31 | 303 | } |
e605f0a7 | 304 | clear_nilfs_sb_dirty(nilfs); |
b2ac86e1 | 305 | return nilfs_sync_super(sbi, flag); |
783f6184 RK |
306 | } |
307 | ||
7ecaa46c RK |
308 | /** |
309 | * nilfs_cleanup_super() - write filesystem state for cleanup | |
310 | * @sbi: nilfs_sb_info to be unmounted or degraded to read-only | |
311 | * | |
312 | * This function restores state flags in the on-disk super block. | |
313 | * This will set "clean" flag (i.e. NILFS_VALID_FS) unless the | |
314 | * filesystem was not clean previously. | |
315 | */ | |
316 | int nilfs_cleanup_super(struct nilfs_sb_info *sbi) | |
317 | { | |
d26493b6 | 318 | struct nilfs_super_block **sbp; |
b2ac86e1 | 319 | int flag = NILFS_SB_COMMIT; |
d26493b6 | 320 | int ret = -EIO; |
7ecaa46c | 321 | |
b2ac86e1 | 322 | sbp = nilfs_prepare_super(sbi, 0); |
d26493b6 JS |
323 | if (sbp) { |
324 | sbp[0]->s_state = cpu_to_le16(sbi->s_nilfs->ns_mount_state); | |
b2ac86e1 JS |
325 | nilfs_set_log_cursor(sbp[0], sbi->s_nilfs); |
326 | if (sbp[1] && sbp[0]->s_last_cno == sbp[1]->s_last_cno) { | |
327 | /* | |
328 | * make the "clean" flag also to the opposite | |
329 | * super block if both super blocks point to | |
330 | * the same checkpoint. | |
331 | */ | |
332 | sbp[1]->s_state = sbp[0]->s_state; | |
333 | flag = NILFS_SB_COMMIT_ALL; | |
334 | } | |
335 | ret = nilfs_commit_super(sbi, flag); | |
d26493b6 | 336 | } |
7ecaa46c RK |
337 | return ret; |
338 | } | |
339 | ||
783f6184 RK |
340 | static void nilfs_put_super(struct super_block *sb) |
341 | { | |
342 | struct nilfs_sb_info *sbi = NILFS_SB(sb); | |
343 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
344 | ||
6cfd0148 CH |
345 | lock_kernel(); |
346 | ||
783f6184 RK |
347 | nilfs_detach_segment_constructor(sbi); |
348 | ||
349 | if (!(sb->s_flags & MS_RDONLY)) { | |
350 | down_write(&nilfs->ns_sem); | |
7ecaa46c | 351 | nilfs_cleanup_super(sbi); |
783f6184 RK |
352 | up_write(&nilfs->ns_sem); |
353 | } | |
e59399d0 | 354 | down_write(&nilfs->ns_super_sem); |
3f82ff55 RK |
355 | if (nilfs->ns_current == sbi) |
356 | nilfs->ns_current = NULL; | |
e59399d0 | 357 | up_write(&nilfs->ns_super_sem); |
783f6184 RK |
358 | |
359 | nilfs_detach_checkpoint(sbi); | |
360 | put_nilfs(sbi->s_nilfs); | |
361 | sbi->s_super = NULL; | |
362 | sb->s_fs_info = NULL; | |
6dd47406 | 363 | nilfs_put_sbinfo(sbi); |
6cfd0148 CH |
364 | |
365 | unlock_kernel(); | |
783f6184 RK |
366 | } |
367 | ||
783f6184 RK |
368 | static int nilfs_sync_fs(struct super_block *sb, int wait) |
369 | { | |
6233caa9 JS |
370 | struct nilfs_sb_info *sbi = NILFS_SB(sb); |
371 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
d26493b6 | 372 | struct nilfs_super_block **sbp; |
783f6184 RK |
373 | int err = 0; |
374 | ||
375 | /* This function is called when super block should be written back */ | |
376 | if (wait) | |
377 | err = nilfs_construct_segment(sb); | |
6233caa9 JS |
378 | |
379 | down_write(&nilfs->ns_sem); | |
d26493b6 | 380 | if (nilfs_sb_dirty(nilfs)) { |
b2ac86e1 JS |
381 | sbp = nilfs_prepare_super(sbi, nilfs_sb_will_flip(nilfs)); |
382 | if (likely(sbp)) { | |
383 | nilfs_set_log_cursor(sbp[0], nilfs); | |
384 | nilfs_commit_super(sbi, NILFS_SB_COMMIT); | |
385 | } | |
d26493b6 | 386 | } |
6233caa9 JS |
387 | up_write(&nilfs->ns_sem); |
388 | ||
783f6184 RK |
389 | return err; |
390 | } | |
391 | ||
392 | int nilfs_attach_checkpoint(struct nilfs_sb_info *sbi, __u64 cno) | |
393 | { | |
394 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
395 | struct nilfs_checkpoint *raw_cp; | |
396 | struct buffer_head *bh_cp; | |
397 | int err; | |
398 | ||
e59399d0 | 399 | down_write(&nilfs->ns_super_sem); |
783f6184 | 400 | list_add(&sbi->s_list, &nilfs->ns_supers); |
e59399d0 | 401 | up_write(&nilfs->ns_super_sem); |
783f6184 | 402 | |
79739565 | 403 | sbi->s_ifile = nilfs_ifile_new(sbi, nilfs->ns_inode_size); |
783f6184 RK |
404 | if (!sbi->s_ifile) |
405 | return -ENOMEM; | |
406 | ||
1154ecbd | 407 | down_read(&nilfs->ns_segctor_sem); |
783f6184 RK |
408 | err = nilfs_cpfile_get_checkpoint(nilfs->ns_cpfile, cno, 0, &raw_cp, |
409 | &bh_cp); | |
1154ecbd | 410 | up_read(&nilfs->ns_segctor_sem); |
783f6184 RK |
411 | if (unlikely(err)) { |
412 | if (err == -ENOENT || err == -EINVAL) { | |
413 | printk(KERN_ERR | |
414 | "NILFS: Invalid checkpoint " | |
415 | "(checkpoint number=%llu)\n", | |
416 | (unsigned long long)cno); | |
417 | err = -EINVAL; | |
418 | } | |
419 | goto failed; | |
420 | } | |
421 | err = nilfs_read_inode_common(sbi->s_ifile, &raw_cp->cp_ifile_inode); | |
422 | if (unlikely(err)) | |
423 | goto failed_bh; | |
424 | atomic_set(&sbi->s_inodes_count, le64_to_cpu(raw_cp->cp_inodes_count)); | |
425 | atomic_set(&sbi->s_blocks_count, le64_to_cpu(raw_cp->cp_blocks_count)); | |
426 | ||
427 | nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, cno, bh_cp); | |
428 | return 0; | |
429 | ||
430 | failed_bh: | |
431 | nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, cno, bh_cp); | |
432 | failed: | |
433 | nilfs_mdt_destroy(sbi->s_ifile); | |
434 | sbi->s_ifile = NULL; | |
435 | ||
e59399d0 | 436 | down_write(&nilfs->ns_super_sem); |
783f6184 | 437 | list_del_init(&sbi->s_list); |
e59399d0 | 438 | up_write(&nilfs->ns_super_sem); |
783f6184 RK |
439 | |
440 | return err; | |
441 | } | |
442 | ||
443 | void nilfs_detach_checkpoint(struct nilfs_sb_info *sbi) | |
444 | { | |
445 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
446 | ||
783f6184 RK |
447 | nilfs_mdt_destroy(sbi->s_ifile); |
448 | sbi->s_ifile = NULL; | |
e59399d0 | 449 | down_write(&nilfs->ns_super_sem); |
783f6184 | 450 | list_del_init(&sbi->s_list); |
e59399d0 | 451 | up_write(&nilfs->ns_super_sem); |
783f6184 RK |
452 | } |
453 | ||
783f6184 RK |
454 | static int nilfs_statfs(struct dentry *dentry, struct kstatfs *buf) |
455 | { | |
456 | struct super_block *sb = dentry->d_sb; | |
457 | struct nilfs_sb_info *sbi = NILFS_SB(sb); | |
c306af23 RK |
458 | struct the_nilfs *nilfs = sbi->s_nilfs; |
459 | u64 id = huge_encode_dev(sb->s_bdev->bd_dev); | |
783f6184 RK |
460 | unsigned long long blocks; |
461 | unsigned long overhead; | |
462 | unsigned long nrsvblocks; | |
463 | sector_t nfreeblocks; | |
783f6184 RK |
464 | int err; |
465 | ||
466 | /* | |
467 | * Compute all of the segment blocks | |
468 | * | |
469 | * The blocks before first segment and after last segment | |
470 | * are excluded. | |
471 | */ | |
472 | blocks = nilfs->ns_blocks_per_segment * nilfs->ns_nsegments | |
473 | - nilfs->ns_first_data_block; | |
474 | nrsvblocks = nilfs->ns_nrsvsegs * nilfs->ns_blocks_per_segment; | |
475 | ||
476 | /* | |
477 | * Compute the overhead | |
478 | * | |
7a65004b | 479 | * When distributing meta data blocks outside segment structure, |
783f6184 RK |
480 | * We must count them as the overhead. |
481 | */ | |
482 | overhead = 0; | |
483 | ||
484 | err = nilfs_count_free_blocks(nilfs, &nfreeblocks); | |
485 | if (unlikely(err)) | |
486 | return err; | |
487 | ||
488 | buf->f_type = NILFS_SUPER_MAGIC; | |
489 | buf->f_bsize = sb->s_blocksize; | |
490 | buf->f_blocks = blocks - overhead; | |
491 | buf->f_bfree = nfreeblocks; | |
492 | buf->f_bavail = (buf->f_bfree >= nrsvblocks) ? | |
493 | (buf->f_bfree - nrsvblocks) : 0; | |
494 | buf->f_files = atomic_read(&sbi->s_inodes_count); | |
495 | buf->f_ffree = 0; /* nilfs_count_free_inodes(sb); */ | |
496 | buf->f_namelen = NILFS_NAME_LEN; | |
c306af23 RK |
497 | buf->f_fsid.val[0] = (u32)id; |
498 | buf->f_fsid.val[1] = (u32)(id >> 32); | |
499 | ||
783f6184 RK |
500 | return 0; |
501 | } | |
502 | ||
b58a285b JS |
503 | static int nilfs_show_options(struct seq_file *seq, struct vfsmount *vfs) |
504 | { | |
505 | struct super_block *sb = vfs->mnt_sb; | |
506 | struct nilfs_sb_info *sbi = NILFS_SB(sb); | |
507 | ||
508 | if (!nilfs_test_opt(sbi, BARRIER)) | |
c6b4d57d | 509 | seq_puts(seq, ",nobarrier"); |
b58a285b JS |
510 | if (nilfs_test_opt(sbi, SNAPSHOT)) |
511 | seq_printf(seq, ",cp=%llu", | |
512 | (unsigned long long int)sbi->s_snapshot_cno); | |
b58a285b | 513 | if (nilfs_test_opt(sbi, ERRORS_PANIC)) |
c6b4d57d | 514 | seq_puts(seq, ",errors=panic"); |
277a6a34 | 515 | if (nilfs_test_opt(sbi, ERRORS_CONT)) |
c6b4d57d | 516 | seq_puts(seq, ",errors=continue"); |
b58a285b | 517 | if (nilfs_test_opt(sbi, STRICT_ORDER)) |
c6b4d57d | 518 | seq_puts(seq, ",order=strict"); |
0234576d | 519 | if (nilfs_test_opt(sbi, NORECOVERY)) |
c6b4d57d | 520 | seq_puts(seq, ",norecovery"); |
e902ec99 | 521 | if (nilfs_test_opt(sbi, DISCARD)) |
c6b4d57d | 522 | seq_puts(seq, ",discard"); |
b58a285b JS |
523 | |
524 | return 0; | |
525 | } | |
526 | ||
b87221de | 527 | static const struct super_operations nilfs_sops = { |
783f6184 RK |
528 | .alloc_inode = nilfs_alloc_inode, |
529 | .destroy_inode = nilfs_destroy_inode, | |
530 | .dirty_inode = nilfs_dirty_inode, | |
531 | /* .write_inode = nilfs_write_inode, */ | |
532 | /* .put_inode = nilfs_put_inode, */ | |
533 | /* .drop_inode = nilfs_drop_inode, */ | |
6fd1e5c9 | 534 | .evict_inode = nilfs_evict_inode, |
783f6184 | 535 | .put_super = nilfs_put_super, |
1dfa2710 | 536 | /* .write_super = nilfs_write_super, */ |
783f6184 RK |
537 | .sync_fs = nilfs_sync_fs, |
538 | /* .write_super_lockfs */ | |
539 | /* .unlockfs */ | |
540 | .statfs = nilfs_statfs, | |
541 | .remount_fs = nilfs_remount, | |
783f6184 | 542 | /* .umount_begin */ |
b58a285b | 543 | .show_options = nilfs_show_options |
783f6184 RK |
544 | }; |
545 | ||
546 | static struct inode * | |
547 | nilfs_nfs_get_inode(struct super_block *sb, u64 ino, u32 generation) | |
548 | { | |
549 | struct inode *inode; | |
550 | ||
551 | if (ino < NILFS_FIRST_INO(sb) && ino != NILFS_ROOT_INO && | |
552 | ino != NILFS_SKETCH_INO) | |
553 | return ERR_PTR(-ESTALE); | |
554 | ||
555 | inode = nilfs_iget(sb, ino); | |
556 | if (IS_ERR(inode)) | |
557 | return ERR_CAST(inode); | |
558 | if (generation && inode->i_generation != generation) { | |
559 | iput(inode); | |
560 | return ERR_PTR(-ESTALE); | |
561 | } | |
562 | ||
563 | return inode; | |
564 | } | |
565 | ||
566 | static struct dentry * | |
567 | nilfs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len, | |
568 | int fh_type) | |
569 | { | |
570 | return generic_fh_to_dentry(sb, fid, fh_len, fh_type, | |
571 | nilfs_nfs_get_inode); | |
572 | } | |
573 | ||
574 | static struct dentry * | |
575 | nilfs_fh_to_parent(struct super_block *sb, struct fid *fid, int fh_len, | |
576 | int fh_type) | |
577 | { | |
578 | return generic_fh_to_parent(sb, fid, fh_len, fh_type, | |
579 | nilfs_nfs_get_inode); | |
580 | } | |
581 | ||
ac4cfdd6 | 582 | static const struct export_operations nilfs_export_ops = { |
783f6184 RK |
583 | .fh_to_dentry = nilfs_fh_to_dentry, |
584 | .fh_to_parent = nilfs_fh_to_parent, | |
585 | .get_parent = nilfs_get_parent, | |
586 | }; | |
587 | ||
588 | enum { | |
589 | Opt_err_cont, Opt_err_panic, Opt_err_ro, | |
773bc4f3 | 590 | Opt_barrier, Opt_nobarrier, Opt_snapshot, Opt_order, Opt_norecovery, |
802d3177 | 591 | Opt_discard, Opt_nodiscard, Opt_err, |
783f6184 RK |
592 | }; |
593 | ||
594 | static match_table_t tokens = { | |
595 | {Opt_err_cont, "errors=continue"}, | |
596 | {Opt_err_panic, "errors=panic"}, | |
597 | {Opt_err_ro, "errors=remount-ro"}, | |
773bc4f3 | 598 | {Opt_barrier, "barrier"}, |
91f1953b | 599 | {Opt_nobarrier, "nobarrier"}, |
783f6184 RK |
600 | {Opt_snapshot, "cp=%u"}, |
601 | {Opt_order, "order=%s"}, | |
0234576d | 602 | {Opt_norecovery, "norecovery"}, |
e902ec99 | 603 | {Opt_discard, "discard"}, |
802d3177 | 604 | {Opt_nodiscard, "nodiscard"}, |
783f6184 RK |
605 | {Opt_err, NULL} |
606 | }; | |
607 | ||
7c017457 | 608 | static int parse_options(char *options, struct super_block *sb, int is_remount) |
783f6184 RK |
609 | { |
610 | struct nilfs_sb_info *sbi = NILFS_SB(sb); | |
611 | char *p; | |
612 | substring_t args[MAX_OPT_ARGS]; | |
613 | int option; | |
614 | ||
615 | if (!options) | |
616 | return 1; | |
617 | ||
618 | while ((p = strsep(&options, ",")) != NULL) { | |
619 | int token; | |
620 | if (!*p) | |
621 | continue; | |
622 | ||
623 | token = match_token(p, tokens, args); | |
624 | switch (token) { | |
773bc4f3 RK |
625 | case Opt_barrier: |
626 | nilfs_set_opt(sbi, BARRIER); | |
627 | break; | |
91f1953b JS |
628 | case Opt_nobarrier: |
629 | nilfs_clear_opt(sbi, BARRIER); | |
783f6184 RK |
630 | break; |
631 | case Opt_order: | |
632 | if (strcmp(args[0].from, "relaxed") == 0) | |
633 | /* Ordered data semantics */ | |
634 | nilfs_clear_opt(sbi, STRICT_ORDER); | |
635 | else if (strcmp(args[0].from, "strict") == 0) | |
636 | /* Strict in-order semantics */ | |
637 | nilfs_set_opt(sbi, STRICT_ORDER); | |
638 | else | |
639 | return 0; | |
640 | break; | |
641 | case Opt_err_panic: | |
642 | nilfs_write_opt(sbi, ERROR_MODE, ERRORS_PANIC); | |
643 | break; | |
644 | case Opt_err_ro: | |
645 | nilfs_write_opt(sbi, ERROR_MODE, ERRORS_RO); | |
646 | break; | |
647 | case Opt_err_cont: | |
648 | nilfs_write_opt(sbi, ERROR_MODE, ERRORS_CONT); | |
649 | break; | |
650 | case Opt_snapshot: | |
651 | if (match_int(&args[0], &option) || option <= 0) | |
652 | return 0; | |
7c017457 RK |
653 | if (is_remount) { |
654 | if (!nilfs_test_opt(sbi, SNAPSHOT)) { | |
655 | printk(KERN_ERR | |
656 | "NILFS: cannot change regular " | |
657 | "mount to snapshot.\n"); | |
658 | return 0; | |
659 | } else if (option != sbi->s_snapshot_cno) { | |
660 | printk(KERN_ERR | |
661 | "NILFS: cannot remount to a " | |
662 | "different snapshot.\n"); | |
663 | return 0; | |
664 | } | |
665 | break; | |
666 | } | |
667 | if (!(sb->s_flags & MS_RDONLY)) { | |
668 | printk(KERN_ERR "NILFS: cannot mount snapshot " | |
669 | "read/write. A read-only option is " | |
670 | "required.\n"); | |
783f6184 | 671 | return 0; |
7c017457 | 672 | } |
783f6184 RK |
673 | sbi->s_snapshot_cno = option; |
674 | nilfs_set_opt(sbi, SNAPSHOT); | |
675 | break; | |
0234576d RK |
676 | case Opt_norecovery: |
677 | nilfs_set_opt(sbi, NORECOVERY); | |
678 | break; | |
e902ec99 JS |
679 | case Opt_discard: |
680 | nilfs_set_opt(sbi, DISCARD); | |
681 | break; | |
802d3177 RK |
682 | case Opt_nodiscard: |
683 | nilfs_clear_opt(sbi, DISCARD); | |
684 | break; | |
783f6184 RK |
685 | default: |
686 | printk(KERN_ERR | |
687 | "NILFS: Unrecognized mount option \"%s\"\n", p); | |
688 | return 0; | |
689 | } | |
690 | } | |
691 | return 1; | |
692 | } | |
693 | ||
694 | static inline void | |
695 | nilfs_set_default_options(struct nilfs_sb_info *sbi, | |
696 | struct nilfs_super_block *sbp) | |
697 | { | |
698 | sbi->s_mount_opt = | |
277a6a34 | 699 | NILFS_MOUNT_ERRORS_RO | NILFS_MOUNT_BARRIER; |
783f6184 RK |
700 | } |
701 | ||
702 | static int nilfs_setup_super(struct nilfs_sb_info *sbi) | |
703 | { | |
704 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
d26493b6 JS |
705 | struct nilfs_super_block **sbp; |
706 | int max_mnt_count; | |
707 | int mnt_count; | |
708 | ||
709 | /* nilfs->ns_sem must be locked by the caller. */ | |
b2ac86e1 | 710 | sbp = nilfs_prepare_super(sbi, 0); |
d26493b6 JS |
711 | if (!sbp) |
712 | return -EIO; | |
713 | ||
714 | max_mnt_count = le16_to_cpu(sbp[0]->s_max_mnt_count); | |
715 | mnt_count = le16_to_cpu(sbp[0]->s_mnt_count); | |
783f6184 | 716 | |
f50a4c81 | 717 | if (nilfs->ns_mount_state & NILFS_ERROR_FS) { |
783f6184 RK |
718 | printk(KERN_WARNING |
719 | "NILFS warning: mounting fs with errors\n"); | |
720 | #if 0 | |
721 | } else if (max_mnt_count >= 0 && mnt_count >= max_mnt_count) { | |
722 | printk(KERN_WARNING | |
723 | "NILFS warning: maximal mount count reached\n"); | |
724 | #endif | |
725 | } | |
726 | if (!max_mnt_count) | |
d26493b6 | 727 | sbp[0]->s_max_mnt_count = cpu_to_le16(NILFS_DFL_MAX_MNT_COUNT); |
783f6184 | 728 | |
d26493b6 JS |
729 | sbp[0]->s_mnt_count = cpu_to_le16(mnt_count + 1); |
730 | sbp[0]->s_state = | |
731 | cpu_to_le16(le16_to_cpu(sbp[0]->s_state) & ~NILFS_VALID_FS); | |
732 | sbp[0]->s_mtime = cpu_to_le64(get_seconds()); | |
b2ac86e1 JS |
733 | /* synchronize sbp[1] with sbp[0] */ |
734 | memcpy(sbp[1], sbp[0], nilfs->ns_sbsize); | |
735 | return nilfs_commit_super(sbi, NILFS_SB_COMMIT_ALL); | |
783f6184 RK |
736 | } |
737 | ||
e339ad31 RK |
738 | struct nilfs_super_block *nilfs_read_super_block(struct super_block *sb, |
739 | u64 pos, int blocksize, | |
740 | struct buffer_head **pbh) | |
783f6184 | 741 | { |
e339ad31 RK |
742 | unsigned long long sb_index = pos; |
743 | unsigned long offset; | |
783f6184 | 744 | |
e339ad31 | 745 | offset = do_div(sb_index, blocksize); |
783f6184 | 746 | *pbh = sb_bread(sb, sb_index); |
e339ad31 | 747 | if (!*pbh) |
783f6184 | 748 | return NULL; |
783f6184 RK |
749 | return (struct nilfs_super_block *)((char *)(*pbh)->b_data + offset); |
750 | } | |
751 | ||
783f6184 RK |
752 | int nilfs_store_magic_and_option(struct super_block *sb, |
753 | struct nilfs_super_block *sbp, | |
754 | char *data) | |
755 | { | |
756 | struct nilfs_sb_info *sbi = NILFS_SB(sb); | |
757 | ||
783f6184 RK |
758 | sb->s_magic = le16_to_cpu(sbp->s_magic); |
759 | ||
760 | /* FS independent flags */ | |
761 | #ifdef NILFS_ATIME_DISABLE | |
762 | sb->s_flags |= MS_NOATIME; | |
763 | #endif | |
764 | ||
783f6184 RK |
765 | nilfs_set_default_options(sbi, sbp); |
766 | ||
767 | sbi->s_resuid = le16_to_cpu(sbp->s_def_resuid); | |
768 | sbi->s_resgid = le16_to_cpu(sbp->s_def_resgid); | |
769 | sbi->s_interval = le32_to_cpu(sbp->s_c_interval); | |
770 | sbi->s_watermark = le32_to_cpu(sbp->s_c_block_max); | |
771 | ||
7c017457 | 772 | return !parse_options(data, sb, 0) ? -EINVAL : 0 ; |
783f6184 RK |
773 | } |
774 | ||
c5ca48aa RK |
775 | int nilfs_check_feature_compatibility(struct super_block *sb, |
776 | struct nilfs_super_block *sbp) | |
777 | { | |
778 | __u64 features; | |
779 | ||
780 | features = le64_to_cpu(sbp->s_feature_incompat) & | |
781 | ~NILFS_FEATURE_INCOMPAT_SUPP; | |
782 | if (features) { | |
783 | printk(KERN_ERR "NILFS: couldn't mount because of unsupported " | |
784 | "optional features (%llx)\n", | |
785 | (unsigned long long)features); | |
786 | return -EINVAL; | |
787 | } | |
788 | features = le64_to_cpu(sbp->s_feature_compat_ro) & | |
789 | ~NILFS_FEATURE_COMPAT_RO_SUPP; | |
790 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
791 | printk(KERN_ERR "NILFS: couldn't mount RDWR because of " | |
792 | "unsupported optional features (%llx)\n", | |
793 | (unsigned long long)features); | |
794 | return -EINVAL; | |
795 | } | |
796 | return 0; | |
797 | } | |
798 | ||
783f6184 RK |
799 | /** |
800 | * nilfs_fill_super() - initialize a super block instance | |
801 | * @sb: super_block | |
802 | * @data: mount options | |
803 | * @silent: silent mode flag | |
804 | * @nilfs: the_nilfs struct | |
805 | * | |
aa7dfb89 | 806 | * This function is called exclusively by nilfs->ns_mount_mutex. |
783f6184 RK |
807 | * So, the recovery process is protected from other simultaneous mounts. |
808 | */ | |
809 | static int | |
810 | nilfs_fill_super(struct super_block *sb, void *data, int silent, | |
811 | struct the_nilfs *nilfs) | |
812 | { | |
813 | struct nilfs_sb_info *sbi; | |
814 | struct inode *root; | |
815 | __u64 cno; | |
816 | int err; | |
817 | ||
818 | sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); | |
819 | if (!sbi) | |
820 | return -ENOMEM; | |
821 | ||
822 | sb->s_fs_info = sbi; | |
823 | ||
824 | get_nilfs(nilfs); | |
825 | sbi->s_nilfs = nilfs; | |
826 | sbi->s_super = sb; | |
6dd47406 | 827 | atomic_set(&sbi->s_count, 1); |
783f6184 RK |
828 | |
829 | err = init_nilfs(nilfs, sbi, (char *)data); | |
830 | if (err) | |
831 | goto failed_sbi; | |
832 | ||
833 | spin_lock_init(&sbi->s_inode_lock); | |
834 | INIT_LIST_HEAD(&sbi->s_dirty_files); | |
835 | INIT_LIST_HEAD(&sbi->s_list); | |
836 | ||
837 | /* | |
838 | * Following initialization is overlapped because | |
839 | * nilfs_sb_info structure has been cleared at the beginning. | |
840 | * But we reserve them to keep our interest and make ready | |
841 | * for the future change. | |
842 | */ | |
843 | get_random_bytes(&sbi->s_next_generation, | |
844 | sizeof(sbi->s_next_generation)); | |
845 | spin_lock_init(&sbi->s_next_gen_lock); | |
846 | ||
847 | sb->s_op = &nilfs_sops; | |
848 | sb->s_export_op = &nilfs_export_ops; | |
849 | sb->s_root = NULL; | |
61239230 | 850 | sb->s_time_gran = 1; |
973bec34 | 851 | sb->s_bdi = nilfs->ns_bdi; |
783f6184 | 852 | |
f50a4c81 RK |
853 | err = load_nilfs(nilfs, sbi); |
854 | if (err) | |
855 | goto failed_sbi; | |
856 | ||
783f6184 RK |
857 | cno = nilfs_last_cno(nilfs); |
858 | ||
859 | if (sb->s_flags & MS_RDONLY) { | |
860 | if (nilfs_test_opt(sbi, SNAPSHOT)) { | |
43be0ec0 | 861 | down_read(&nilfs->ns_segctor_sem); |
1f5abe7e RK |
862 | err = nilfs_cpfile_is_snapshot(nilfs->ns_cpfile, |
863 | sbi->s_snapshot_cno); | |
43be0ec0 ZY |
864 | up_read(&nilfs->ns_segctor_sem); |
865 | if (err < 0) { | |
866 | if (err == -ENOENT) | |
867 | err = -EINVAL; | |
1f5abe7e | 868 | goto failed_sbi; |
43be0ec0 | 869 | } |
1f5abe7e | 870 | if (!err) { |
783f6184 RK |
871 | printk(KERN_ERR |
872 | "NILFS: The specified checkpoint is " | |
873 | "not a snapshot " | |
874 | "(checkpoint number=%llu).\n", | |
875 | (unsigned long long)sbi->s_snapshot_cno); | |
876 | err = -EINVAL; | |
877 | goto failed_sbi; | |
878 | } | |
879 | cno = sbi->s_snapshot_cno; | |
d240e067 | 880 | } |
783f6184 RK |
881 | } |
882 | ||
883 | err = nilfs_attach_checkpoint(sbi, cno); | |
884 | if (err) { | |
885 | printk(KERN_ERR "NILFS: error loading a checkpoint" | |
886 | " (checkpoint number=%llu).\n", (unsigned long long)cno); | |
887 | goto failed_sbi; | |
888 | } | |
889 | ||
890 | if (!(sb->s_flags & MS_RDONLY)) { | |
cece5520 | 891 | err = nilfs_attach_segment_constructor(sbi); |
783f6184 RK |
892 | if (err) |
893 | goto failed_checkpoint; | |
894 | } | |
895 | ||
896 | root = nilfs_iget(sb, NILFS_ROOT_INO); | |
897 | if (IS_ERR(root)) { | |
898 | printk(KERN_ERR "NILFS: get root inode failed\n"); | |
899 | err = PTR_ERR(root); | |
900 | goto failed_segctor; | |
901 | } | |
902 | if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { | |
903 | iput(root); | |
904 | printk(KERN_ERR "NILFS: corrupt root inode.\n"); | |
905 | err = -EINVAL; | |
906 | goto failed_segctor; | |
907 | } | |
908 | sb->s_root = d_alloc_root(root); | |
909 | if (!sb->s_root) { | |
910 | iput(root); | |
911 | printk(KERN_ERR "NILFS: get root dentry failed\n"); | |
912 | err = -ENOMEM; | |
913 | goto failed_segctor; | |
914 | } | |
915 | ||
916 | if (!(sb->s_flags & MS_RDONLY)) { | |
917 | down_write(&nilfs->ns_sem); | |
918 | nilfs_setup_super(sbi); | |
919 | up_write(&nilfs->ns_sem); | |
920 | } | |
921 | ||
e59399d0 | 922 | down_write(&nilfs->ns_super_sem); |
3f82ff55 RK |
923 | if (!nilfs_test_opt(sbi, SNAPSHOT)) |
924 | nilfs->ns_current = sbi; | |
e59399d0 | 925 | up_write(&nilfs->ns_super_sem); |
3f82ff55 | 926 | |
783f6184 RK |
927 | return 0; |
928 | ||
783f6184 RK |
929 | failed_segctor: |
930 | nilfs_detach_segment_constructor(sbi); | |
931 | ||
932 | failed_checkpoint: | |
933 | nilfs_detach_checkpoint(sbi); | |
934 | ||
935 | failed_sbi: | |
936 | put_nilfs(nilfs); | |
937 | sb->s_fs_info = NULL; | |
6dd47406 | 938 | nilfs_put_sbinfo(sbi); |
783f6184 RK |
939 | return err; |
940 | } | |
941 | ||
942 | static int nilfs_remount(struct super_block *sb, int *flags, char *data) | |
943 | { | |
944 | struct nilfs_sb_info *sbi = NILFS_SB(sb); | |
783f6184 RK |
945 | struct the_nilfs *nilfs = sbi->s_nilfs; |
946 | unsigned long old_sb_flags; | |
947 | struct nilfs_mount_options old_opts; | |
d240e067 | 948 | int was_snapshot, err; |
783f6184 | 949 | |
337eb00a AIB |
950 | lock_kernel(); |
951 | ||
e59399d0 | 952 | down_write(&nilfs->ns_super_sem); |
783f6184 RK |
953 | old_sb_flags = sb->s_flags; |
954 | old_opts.mount_opt = sbi->s_mount_opt; | |
955 | old_opts.snapshot_cno = sbi->s_snapshot_cno; | |
d240e067 | 956 | was_snapshot = nilfs_test_opt(sbi, SNAPSHOT); |
783f6184 | 957 | |
7c017457 | 958 | if (!parse_options(data, sb, 1)) { |
783f6184 RK |
959 | err = -EINVAL; |
960 | goto restore_opts; | |
961 | } | |
962 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL); | |
963 | ||
d240e067 | 964 | err = -EINVAL; |
7c017457 RK |
965 | if (was_snapshot && !(*flags & MS_RDONLY)) { |
966 | printk(KERN_ERR "NILFS (device %s): cannot remount snapshot " | |
967 | "read/write.\n", sb->s_id); | |
968 | goto restore_opts; | |
783f6184 RK |
969 | } |
970 | ||
0234576d RK |
971 | if (!nilfs_valid_fs(nilfs)) { |
972 | printk(KERN_WARNING "NILFS (device %s): couldn't " | |
973 | "remount because the filesystem is in an " | |
974 | "incomplete recovery state.\n", sb->s_id); | |
0234576d RK |
975 | goto restore_opts; |
976 | } | |
977 | ||
783f6184 RK |
978 | if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) |
979 | goto out; | |
980 | if (*flags & MS_RDONLY) { | |
981 | /* Shutting down the segment constructor */ | |
982 | nilfs_detach_segment_constructor(sbi); | |
983 | sb->s_flags |= MS_RDONLY; | |
984 | ||
783f6184 RK |
985 | /* |
986 | * Remounting a valid RW partition RDONLY, so set | |
987 | * the RDONLY flag and then mark the partition as valid again. | |
988 | */ | |
989 | down_write(&nilfs->ns_sem); | |
7ecaa46c | 990 | nilfs_cleanup_super(sbi); |
783f6184 RK |
991 | up_write(&nilfs->ns_sem); |
992 | } else { | |
c5ca48aa RK |
993 | __u64 features; |
994 | ||
783f6184 RK |
995 | /* |
996 | * Mounting a RDONLY partition read-write, so reread and | |
997 | * store the current valid flag. (It may have been changed | |
998 | * by fsck since we originally mounted the partition.) | |
999 | */ | |
c5ca48aa RK |
1000 | down_read(&nilfs->ns_sem); |
1001 | features = le64_to_cpu(nilfs->ns_sbp[0]->s_feature_compat_ro) & | |
1002 | ~NILFS_FEATURE_COMPAT_RO_SUPP; | |
1003 | up_read(&nilfs->ns_sem); | |
1004 | if (features) { | |
1005 | printk(KERN_WARNING "NILFS (device %s): couldn't " | |
1006 | "remount RDWR because of unsupported optional " | |
1007 | "features (%llx)\n", | |
1008 | sb->s_id, (unsigned long long)features); | |
1009 | err = -EROFS; | |
1010 | goto restore_opts; | |
1011 | } | |
1012 | ||
783f6184 | 1013 | sb->s_flags &= ~MS_RDONLY; |
783f6184 | 1014 | |
cece5520 | 1015 | err = nilfs_attach_segment_constructor(sbi); |
783f6184 | 1016 | if (err) |
e59399d0 | 1017 | goto restore_opts; |
783f6184 RK |
1018 | |
1019 | down_write(&nilfs->ns_sem); | |
1020 | nilfs_setup_super(sbi); | |
1021 | up_write(&nilfs->ns_sem); | |
783f6184 RK |
1022 | } |
1023 | out: | |
e59399d0 | 1024 | up_write(&nilfs->ns_super_sem); |
337eb00a | 1025 | unlock_kernel(); |
783f6184 RK |
1026 | return 0; |
1027 | ||
783f6184 RK |
1028 | restore_opts: |
1029 | sb->s_flags = old_sb_flags; | |
1030 | sbi->s_mount_opt = old_opts.mount_opt; | |
1031 | sbi->s_snapshot_cno = old_opts.snapshot_cno; | |
e59399d0 | 1032 | up_write(&nilfs->ns_super_sem); |
337eb00a | 1033 | unlock_kernel(); |
783f6184 RK |
1034 | return err; |
1035 | } | |
1036 | ||
1037 | struct nilfs_super_data { | |
1038 | struct block_device *bdev; | |
6dd47406 | 1039 | struct nilfs_sb_info *sbi; |
783f6184 RK |
1040 | __u64 cno; |
1041 | int flags; | |
1042 | }; | |
1043 | ||
1044 | /** | |
1045 | * nilfs_identify - pre-read mount options needed to identify mount instance | |
1046 | * @data: mount options | |
1047 | * @sd: nilfs_super_data | |
1048 | */ | |
1049 | static int nilfs_identify(char *data, struct nilfs_super_data *sd) | |
1050 | { | |
1051 | char *p, *options = data; | |
1052 | substring_t args[MAX_OPT_ARGS]; | |
1053 | int option, token; | |
1054 | int ret = 0; | |
1055 | ||
1056 | do { | |
1057 | p = strsep(&options, ","); | |
1058 | if (p != NULL && *p) { | |
1059 | token = match_token(p, tokens, args); | |
1060 | if (token == Opt_snapshot) { | |
1061 | if (!(sd->flags & MS_RDONLY)) | |
1062 | ret++; | |
1063 | else { | |
1064 | ret = match_int(&args[0], &option); | |
1065 | if (!ret) { | |
1066 | if (option > 0) | |
1067 | sd->cno = option; | |
1068 | else | |
1069 | ret++; | |
1070 | } | |
1071 | } | |
1072 | } | |
1073 | if (ret) | |
1074 | printk(KERN_ERR | |
1075 | "NILFS: invalid mount option: %s\n", p); | |
1076 | } | |
1077 | if (!options) | |
1078 | break; | |
1079 | BUG_ON(options == data); | |
1080 | *(options - 1) = ','; | |
1081 | } while (!ret); | |
1082 | return ret; | |
1083 | } | |
1084 | ||
1085 | static int nilfs_set_bdev_super(struct super_block *s, void *data) | |
1086 | { | |
1087 | struct nilfs_super_data *sd = data; | |
1088 | ||
1089 | s->s_bdev = sd->bdev; | |
1090 | s->s_dev = s->s_bdev->bd_dev; | |
1091 | return 0; | |
1092 | } | |
1093 | ||
1094 | static int nilfs_test_bdev_super(struct super_block *s, void *data) | |
783f6184 RK |
1095 | { |
1096 | struct nilfs_super_data *sd = data; | |
6dd47406 RK |
1097 | |
1098 | return sd->sbi && s->s_fs_info == (void *)sd->sbi; | |
783f6184 RK |
1099 | } |
1100 | ||
1101 | static int | |
1102 | nilfs_get_sb(struct file_system_type *fs_type, int flags, | |
1103 | const char *dev_name, void *data, struct vfsmount *mnt) | |
1104 | { | |
1105 | struct nilfs_super_data sd; | |
33c8e57c | 1106 | struct super_block *s; |
13e90559 | 1107 | fmode_t mode = FMODE_READ; |
33c8e57c | 1108 | struct the_nilfs *nilfs; |
783f6184 RK |
1109 | int err, need_to_close = 1; |
1110 | ||
13e90559 RK |
1111 | if (!(flags & MS_RDONLY)) |
1112 | mode |= FMODE_WRITE; | |
1113 | ||
1114 | sd.bdev = open_bdev_exclusive(dev_name, mode, fs_type); | |
783f6184 RK |
1115 | if (IS_ERR(sd.bdev)) |
1116 | return PTR_ERR(sd.bdev); | |
1117 | ||
1118 | /* | |
1119 | * To get mount instance using sget() vfs-routine, NILFS needs | |
1120 | * much more information than normal filesystems to identify mount | |
1121 | * instance. For snapshot mounts, not only a mount type (ro-mount | |
1122 | * or rw-mount) but also a checkpoint number is required. | |
783f6184 RK |
1123 | */ |
1124 | sd.cno = 0; | |
1125 | sd.flags = flags; | |
1126 | if (nilfs_identify((char *)data, &sd)) { | |
1127 | err = -EINVAL; | |
1128 | goto failed; | |
1129 | } | |
1130 | ||
33c8e57c RK |
1131 | nilfs = find_or_create_nilfs(sd.bdev); |
1132 | if (!nilfs) { | |
1133 | err = -ENOMEM; | |
1134 | goto failed; | |
1135 | } | |
1136 | ||
aa7dfb89 | 1137 | mutex_lock(&nilfs->ns_mount_mutex); |
3f82ff55 RK |
1138 | |
1139 | if (!sd.cno) { | |
1140 | /* | |
1141 | * Check if an exclusive mount exists or not. | |
1142 | * Snapshot mounts coexist with a current mount | |
1143 | * (i.e. rw-mount or ro-mount), whereas rw-mount and | |
1144 | * ro-mount are mutually exclusive. | |
1145 | */ | |
e59399d0 | 1146 | down_read(&nilfs->ns_super_sem); |
3f82ff55 RK |
1147 | if (nilfs->ns_current && |
1148 | ((nilfs->ns_current->s_super->s_flags ^ flags) | |
1149 | & MS_RDONLY)) { | |
e59399d0 | 1150 | up_read(&nilfs->ns_super_sem); |
3f82ff55 RK |
1151 | err = -EBUSY; |
1152 | goto failed_unlock; | |
1153 | } | |
e59399d0 | 1154 | up_read(&nilfs->ns_super_sem); |
783f6184 RK |
1155 | } |
1156 | ||
1157 | /* | |
6dd47406 | 1158 | * Find existing nilfs_sb_info struct |
783f6184 | 1159 | */ |
6dd47406 RK |
1160 | sd.sbi = nilfs_find_sbinfo(nilfs, !(flags & MS_RDONLY), sd.cno); |
1161 | ||
6dd47406 RK |
1162 | /* |
1163 | * Get super block instance holding the nilfs_sb_info struct. | |
1164 | * A new instance is allocated if no existing mount is present or | |
1165 | * existing instance has been unmounted. | |
1166 | */ | |
33c8e57c | 1167 | s = sget(fs_type, nilfs_test_bdev_super, nilfs_set_bdev_super, &sd); |
6dd47406 RK |
1168 | if (sd.sbi) |
1169 | nilfs_put_sbinfo(sd.sbi); | |
1170 | ||
33c8e57c RK |
1171 | if (IS_ERR(s)) { |
1172 | err = PTR_ERR(s); | |
1173 | goto failed_unlock; | |
783f6184 RK |
1174 | } |
1175 | ||
1176 | if (!s->s_root) { | |
1177 | char b[BDEVNAME_SIZE]; | |
1178 | ||
33c8e57c | 1179 | /* New superblock instance created */ |
783f6184 | 1180 | s->s_flags = flags; |
4571b82c | 1181 | s->s_mode = mode; |
783f6184 RK |
1182 | strlcpy(s->s_id, bdevname(sd.bdev, b), sizeof(s->s_id)); |
1183 | sb_set_blocksize(s, block_size(sd.bdev)); | |
1184 | ||
e2d1591a RK |
1185 | err = nilfs_fill_super(s, data, flags & MS_SILENT ? 1 : 0, |
1186 | nilfs); | |
783f6184 RK |
1187 | if (err) |
1188 | goto cancel_new; | |
1189 | ||
1190 | s->s_flags |= MS_ACTIVE; | |
1191 | need_to_close = 0; | |
783f6184 RK |
1192 | } |
1193 | ||
aa7dfb89 | 1194 | mutex_unlock(&nilfs->ns_mount_mutex); |
783f6184 RK |
1195 | put_nilfs(nilfs); |
1196 | if (need_to_close) | |
13e90559 | 1197 | close_bdev_exclusive(sd.bdev, mode); |
783f6184 RK |
1198 | simple_set_mnt(mnt, s); |
1199 | return 0; | |
1200 | ||
783f6184 | 1201 | failed_unlock: |
aa7dfb89 | 1202 | mutex_unlock(&nilfs->ns_mount_mutex); |
33c8e57c | 1203 | put_nilfs(nilfs); |
783f6184 | 1204 | failed: |
13e90559 | 1205 | close_bdev_exclusive(sd.bdev, mode); |
783f6184 RK |
1206 | |
1207 | return err; | |
1208 | ||
1209 | cancel_new: | |
1210 | /* Abandoning the newly allocated superblock */ | |
aa7dfb89 | 1211 | mutex_unlock(&nilfs->ns_mount_mutex); |
33c8e57c | 1212 | put_nilfs(nilfs); |
a95161aa | 1213 | deactivate_locked_super(s); |
783f6184 | 1214 | /* |
b87ca919 | 1215 | * deactivate_locked_super() invokes close_bdev_exclusive(). |
783f6184 | 1216 | * We must finish all post-cleaning before this call; |
aa7dfb89 | 1217 | * put_nilfs() needs the block device. |
783f6184 RK |
1218 | */ |
1219 | return err; | |
1220 | } | |
1221 | ||
783f6184 RK |
1222 | struct file_system_type nilfs_fs_type = { |
1223 | .owner = THIS_MODULE, | |
1224 | .name = "nilfs2", | |
1225 | .get_sb = nilfs_get_sb, | |
1226 | .kill_sb = kill_block_super, | |
1227 | .fs_flags = FS_REQUIRES_DEV, | |
1228 | }; | |
1229 | ||
41c88bd7 | 1230 | static void nilfs_inode_init_once(void *obj) |
783f6184 | 1231 | { |
41c88bd7 | 1232 | struct nilfs_inode_info *ii = obj; |
783f6184 | 1233 | |
41c88bd7 LH |
1234 | INIT_LIST_HEAD(&ii->i_dirty); |
1235 | #ifdef CONFIG_NILFS_XATTR | |
1236 | init_rwsem(&ii->xattr_sem); | |
1237 | #endif | |
1238 | nilfs_btnode_cache_init_once(&ii->i_btnode_cache); | |
05d0e94b | 1239 | ii->i_bmap = &ii->i_bmap_data; |
41c88bd7 LH |
1240 | inode_init_once(&ii->vfs_inode); |
1241 | } | |
783f6184 | 1242 | |
41c88bd7 LH |
1243 | static void nilfs_segbuf_init_once(void *obj) |
1244 | { | |
1245 | memset(obj, 0, sizeof(struct nilfs_segment_buffer)); | |
1246 | } | |
783f6184 | 1247 | |
41c88bd7 LH |
1248 | static void nilfs_destroy_cachep(void) |
1249 | { | |
84cb0999 | 1250 | if (nilfs_inode_cachep) |
41c88bd7 | 1251 | kmem_cache_destroy(nilfs_inode_cachep); |
84cb0999 | 1252 | if (nilfs_transaction_cachep) |
41c88bd7 | 1253 | kmem_cache_destroy(nilfs_transaction_cachep); |
84cb0999 | 1254 | if (nilfs_segbuf_cachep) |
41c88bd7 | 1255 | kmem_cache_destroy(nilfs_segbuf_cachep); |
84cb0999 | 1256 | if (nilfs_btree_path_cache) |
41c88bd7 LH |
1257 | kmem_cache_destroy(nilfs_btree_path_cache); |
1258 | } | |
783f6184 | 1259 | |
41c88bd7 LH |
1260 | static int __init nilfs_init_cachep(void) |
1261 | { | |
1262 | nilfs_inode_cachep = kmem_cache_create("nilfs2_inode_cache", | |
1263 | sizeof(struct nilfs_inode_info), 0, | |
1264 | SLAB_RECLAIM_ACCOUNT, nilfs_inode_init_once); | |
1265 | if (!nilfs_inode_cachep) | |
1266 | goto fail; | |
1267 | ||
1268 | nilfs_transaction_cachep = kmem_cache_create("nilfs2_transaction_cache", | |
1269 | sizeof(struct nilfs_transaction_info), 0, | |
1270 | SLAB_RECLAIM_ACCOUNT, NULL); | |
1271 | if (!nilfs_transaction_cachep) | |
1272 | goto fail; | |
1273 | ||
1274 | nilfs_segbuf_cachep = kmem_cache_create("nilfs2_segbuf_cache", | |
1275 | sizeof(struct nilfs_segment_buffer), 0, | |
1276 | SLAB_RECLAIM_ACCOUNT, nilfs_segbuf_init_once); | |
1277 | if (!nilfs_segbuf_cachep) | |
1278 | goto fail; | |
1279 | ||
1280 | nilfs_btree_path_cache = kmem_cache_create("nilfs2_btree_path_cache", | |
1281 | sizeof(struct nilfs_btree_path) * NILFS_BTREE_LEVEL_MAX, | |
1282 | 0, 0, NULL); | |
1283 | if (!nilfs_btree_path_cache) | |
1284 | goto fail; | |
783f6184 RK |
1285 | |
1286 | return 0; | |
1287 | ||
41c88bd7 LH |
1288 | fail: |
1289 | nilfs_destroy_cachep(); | |
1290 | return -ENOMEM; | |
1291 | } | |
1292 | ||
1293 | static int __init init_nilfs_fs(void) | |
1294 | { | |
1295 | int err; | |
783f6184 | 1296 | |
41c88bd7 LH |
1297 | err = nilfs_init_cachep(); |
1298 | if (err) | |
1299 | goto fail; | |
783f6184 | 1300 | |
41c88bd7 LH |
1301 | err = register_filesystem(&nilfs_fs_type); |
1302 | if (err) | |
1303 | goto free_cachep; | |
783f6184 | 1304 | |
9f130263 | 1305 | printk(KERN_INFO "NILFS version 2 loaded\n"); |
41c88bd7 | 1306 | return 0; |
783f6184 | 1307 | |
41c88bd7 LH |
1308 | free_cachep: |
1309 | nilfs_destroy_cachep(); | |
1310 | fail: | |
783f6184 RK |
1311 | return err; |
1312 | } | |
1313 | ||
1314 | static void __exit exit_nilfs_fs(void) | |
1315 | { | |
41c88bd7 | 1316 | nilfs_destroy_cachep(); |
783f6184 RK |
1317 | unregister_filesystem(&nilfs_fs_type); |
1318 | } | |
1319 | ||
1320 | module_init(init_nilfs_fs) | |
1321 | module_exit(exit_nilfs_fs) |