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> | |
53 | #include "nilfs.h" | |
54 | #include "mdt.h" | |
55 | #include "alloc.h" | |
56 | #include "page.h" | |
57 | #include "cpfile.h" | |
58 | #include "ifile.h" | |
59 | #include "dat.h" | |
60 | #include "segment.h" | |
61 | #include "segbuf.h" | |
62 | ||
63 | MODULE_AUTHOR("NTT Corp."); | |
64 | MODULE_DESCRIPTION("A New Implementation of the Log-structured Filesystem " | |
65 | "(NILFS)"); | |
783f6184 RK |
66 | MODULE_LICENSE("GPL"); |
67 | ||
8c85e125 | 68 | static void nilfs_write_super(struct super_block *sb); |
783f6184 RK |
69 | static int nilfs_remount(struct super_block *sb, int *flags, char *data); |
70 | static int test_exclusive_mount(struct file_system_type *fs_type, | |
71 | struct block_device *bdev, int flags); | |
72 | ||
73 | /** | |
74 | * nilfs_error() - report failure condition on a filesystem | |
75 | * | |
76 | * nilfs_error() sets an ERROR_FS flag on the superblock as well as | |
77 | * reporting an error message. It should be called when NILFS detects | |
78 | * incoherences or defects of meta data on disk. As for sustainable | |
79 | * errors such as a single-shot I/O error, nilfs_warning() or the printk() | |
80 | * function should be used instead. | |
81 | * | |
82 | * The segment constructor must not call this function because it can | |
83 | * kill itself. | |
84 | */ | |
85 | void nilfs_error(struct super_block *sb, const char *function, | |
86 | const char *fmt, ...) | |
87 | { | |
88 | struct nilfs_sb_info *sbi = NILFS_SB(sb); | |
89 | va_list args; | |
90 | ||
91 | va_start(args, fmt); | |
92 | printk(KERN_CRIT "NILFS error (device %s): %s: ", sb->s_id, function); | |
93 | vprintk(fmt, args); | |
94 | printk("\n"); | |
95 | va_end(args); | |
96 | ||
97 | if (!(sb->s_flags & MS_RDONLY)) { | |
98 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
99 | ||
100 | if (!nilfs_test_opt(sbi, ERRORS_CONT)) | |
101 | nilfs_detach_segment_constructor(sbi); | |
102 | ||
103 | down_write(&nilfs->ns_sem); | |
104 | if (!(nilfs->ns_mount_state & NILFS_ERROR_FS)) { | |
105 | nilfs->ns_mount_state |= NILFS_ERROR_FS; | |
e339ad31 RK |
106 | nilfs->ns_sbp[0]->s_state |= |
107 | cpu_to_le16(NILFS_ERROR_FS); | |
108 | nilfs_commit_super(sbi, 1); | |
783f6184 RK |
109 | } |
110 | up_write(&nilfs->ns_sem); | |
111 | ||
112 | if (nilfs_test_opt(sbi, ERRORS_RO)) { | |
113 | printk(KERN_CRIT "Remounting filesystem read-only\n"); | |
114 | sb->s_flags |= MS_RDONLY; | |
115 | } | |
116 | } | |
117 | ||
118 | if (nilfs_test_opt(sbi, ERRORS_PANIC)) | |
119 | panic("NILFS (device %s): panic forced after error\n", | |
120 | sb->s_id); | |
121 | } | |
122 | ||
123 | void nilfs_warning(struct super_block *sb, const char *function, | |
124 | const char *fmt, ...) | |
125 | { | |
126 | va_list args; | |
127 | ||
128 | va_start(args, fmt); | |
129 | printk(KERN_WARNING "NILFS warning (device %s): %s: ", | |
130 | sb->s_id, function); | |
131 | vprintk(fmt, args); | |
132 | printk("\n"); | |
133 | va_end(args); | |
134 | } | |
135 | ||
136 | static struct kmem_cache *nilfs_inode_cachep; | |
137 | ||
138 | struct inode *nilfs_alloc_inode(struct super_block *sb) | |
139 | { | |
140 | struct nilfs_inode_info *ii; | |
141 | ||
142 | ii = kmem_cache_alloc(nilfs_inode_cachep, GFP_NOFS); | |
143 | if (!ii) | |
144 | return NULL; | |
145 | ii->i_bh = NULL; | |
146 | ii->i_state = 0; | |
147 | ii->vfs_inode.i_version = 1; | |
148 | nilfs_btnode_cache_init(&ii->i_btnode_cache); | |
149 | return &ii->vfs_inode; | |
150 | } | |
151 | ||
152 | void nilfs_destroy_inode(struct inode *inode) | |
153 | { | |
154 | kmem_cache_free(nilfs_inode_cachep, NILFS_I(inode)); | |
155 | } | |
156 | ||
157 | static void init_once(void *obj) | |
158 | { | |
159 | struct nilfs_inode_info *ii = obj; | |
160 | ||
161 | INIT_LIST_HEAD(&ii->i_dirty); | |
162 | #ifdef CONFIG_NILFS_XATTR | |
163 | init_rwsem(&ii->xattr_sem); | |
164 | #endif | |
165 | nilfs_btnode_cache_init_once(&ii->i_btnode_cache); | |
166 | ii->i_bmap = (struct nilfs_bmap *)&ii->i_bmap_union; | |
167 | inode_init_once(&ii->vfs_inode); | |
168 | } | |
169 | ||
170 | static int nilfs_init_inode_cache(void) | |
171 | { | |
172 | nilfs_inode_cachep = kmem_cache_create("nilfs2_inode_cache", | |
173 | sizeof(struct nilfs_inode_info), | |
174 | 0, SLAB_RECLAIM_ACCOUNT, | |
175 | init_once); | |
176 | ||
177 | return (nilfs_inode_cachep == NULL) ? -ENOMEM : 0; | |
178 | } | |
179 | ||
180 | static inline void nilfs_destroy_inode_cache(void) | |
181 | { | |
182 | kmem_cache_destroy(nilfs_inode_cachep); | |
183 | } | |
184 | ||
185 | static void nilfs_clear_inode(struct inode *inode) | |
186 | { | |
187 | struct nilfs_inode_info *ii = NILFS_I(inode); | |
783f6184 RK |
188 | |
189 | #ifdef CONFIG_NILFS_POSIX_ACL | |
190 | if (ii->i_acl && ii->i_acl != NILFS_ACL_NOT_CACHED) { | |
191 | posix_acl_release(ii->i_acl); | |
192 | ii->i_acl = NILFS_ACL_NOT_CACHED; | |
193 | } | |
194 | if (ii->i_default_acl && ii->i_default_acl != NILFS_ACL_NOT_CACHED) { | |
195 | posix_acl_release(ii->i_default_acl); | |
196 | ii->i_default_acl = NILFS_ACL_NOT_CACHED; | |
197 | } | |
198 | #endif | |
199 | /* | |
200 | * Free resources allocated in nilfs_read_inode(), here. | |
201 | */ | |
a2e7d2df | 202 | BUG_ON(!list_empty(&ii->i_dirty)); |
783f6184 RK |
203 | brelse(ii->i_bh); |
204 | ii->i_bh = NULL; | |
783f6184 RK |
205 | |
206 | if (test_bit(NILFS_I_BMAP, &ii->i_state)) | |
207 | nilfs_bmap_clear(ii->i_bmap); | |
208 | ||
209 | nilfs_btnode_cache_clear(&ii->i_btnode_cache); | |
783f6184 RK |
210 | } |
211 | ||
e339ad31 | 212 | static int nilfs_sync_super(struct nilfs_sb_info *sbi, int dupsb) |
783f6184 RK |
213 | { |
214 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
215 | int err; | |
216 | int barrier_done = 0; | |
217 | ||
218 | if (nilfs_test_opt(sbi, BARRIER)) { | |
e339ad31 | 219 | set_buffer_ordered(nilfs->ns_sbh[0]); |
783f6184 RK |
220 | barrier_done = 1; |
221 | } | |
222 | retry: | |
e339ad31 RK |
223 | set_buffer_dirty(nilfs->ns_sbh[0]); |
224 | err = sync_dirty_buffer(nilfs->ns_sbh[0]); | |
783f6184 RK |
225 | if (err == -EOPNOTSUPP && barrier_done) { |
226 | nilfs_warning(sbi->s_super, __func__, | |
227 | "barrier-based sync failed. " | |
228 | "disabling barriers\n"); | |
229 | nilfs_clear_opt(sbi, BARRIER); | |
230 | barrier_done = 0; | |
e339ad31 | 231 | clear_buffer_ordered(nilfs->ns_sbh[0]); |
783f6184 RK |
232 | goto retry; |
233 | } | |
e339ad31 | 234 | if (unlikely(err)) { |
783f6184 RK |
235 | printk(KERN_ERR |
236 | "NILFS: unable to write superblock (err=%d)\n", err); | |
e339ad31 RK |
237 | if (err == -EIO && nilfs->ns_sbh[1]) { |
238 | nilfs_fall_back_super_block(nilfs); | |
239 | goto retry; | |
240 | } | |
241 | } else { | |
242 | struct nilfs_super_block *sbp = nilfs->ns_sbp[0]; | |
243 | ||
244 | /* | |
245 | * The latest segment becomes trailable from the position | |
246 | * written in superblock. | |
247 | */ | |
783f6184 | 248 | clear_nilfs_discontinued(nilfs); |
e339ad31 RK |
249 | |
250 | /* update GC protection for recent segments */ | |
251 | if (nilfs->ns_sbh[1]) { | |
252 | sbp = NULL; | |
253 | if (dupsb) { | |
254 | set_buffer_dirty(nilfs->ns_sbh[1]); | |
255 | if (!sync_dirty_buffer(nilfs->ns_sbh[1])) | |
256 | sbp = nilfs->ns_sbp[1]; | |
257 | } | |
258 | } | |
259 | if (sbp) { | |
260 | spin_lock(&nilfs->ns_last_segment_lock); | |
261 | nilfs->ns_prot_seq = le64_to_cpu(sbp->s_last_seq); | |
262 | spin_unlock(&nilfs->ns_last_segment_lock); | |
263 | } | |
783f6184 RK |
264 | } |
265 | ||
266 | return err; | |
267 | } | |
268 | ||
e339ad31 | 269 | int nilfs_commit_super(struct nilfs_sb_info *sbi, int dupsb) |
783f6184 RK |
270 | { |
271 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
e339ad31 | 272 | struct nilfs_super_block **sbp = nilfs->ns_sbp; |
783f6184 | 273 | sector_t nfreeblocks; |
e339ad31 | 274 | time_t t; |
783f6184 RK |
275 | int err; |
276 | ||
277 | /* nilfs->sem must be locked by the caller. */ | |
e339ad31 RK |
278 | if (sbp[0]->s_magic != NILFS_SUPER_MAGIC) { |
279 | if (sbp[1] && sbp[1]->s_magic == NILFS_SUPER_MAGIC) | |
280 | nilfs_swap_super_block(nilfs); | |
281 | else { | |
282 | printk(KERN_CRIT "NILFS: superblock broke on dev %s\n", | |
283 | sbi->s_super->s_id); | |
284 | return -EIO; | |
285 | } | |
286 | } | |
783f6184 RK |
287 | err = nilfs_count_free_blocks(nilfs, &nfreeblocks); |
288 | if (unlikely(err)) { | |
289 | printk(KERN_ERR "NILFS: failed to count free blocks\n"); | |
290 | return err; | |
291 | } | |
e339ad31 RK |
292 | spin_lock(&nilfs->ns_last_segment_lock); |
293 | sbp[0]->s_last_seq = cpu_to_le64(nilfs->ns_last_seq); | |
294 | sbp[0]->s_last_pseg = cpu_to_le64(nilfs->ns_last_pseg); | |
295 | sbp[0]->s_last_cno = cpu_to_le64(nilfs->ns_last_cno); | |
296 | spin_unlock(&nilfs->ns_last_segment_lock); | |
297 | ||
298 | t = get_seconds(); | |
299 | nilfs->ns_sbwtime[0] = t; | |
300 | sbp[0]->s_free_blocks_count = cpu_to_le64(nfreeblocks); | |
301 | sbp[0]->s_wtime = cpu_to_le64(t); | |
302 | sbp[0]->s_sum = 0; | |
303 | sbp[0]->s_sum = cpu_to_le32(crc32_le(nilfs->ns_crc_seed, | |
304 | (unsigned char *)sbp[0], | |
305 | nilfs->ns_sbsize)); | |
306 | if (dupsb && sbp[1]) { | |
307 | memcpy(sbp[1], sbp[0], nilfs->ns_sbsize); | |
308 | nilfs->ns_sbwtime[1] = t; | |
309 | } | |
783f6184 | 310 | sbi->s_super->s_dirt = 0; |
e339ad31 | 311 | return nilfs_sync_super(sbi, dupsb); |
783f6184 RK |
312 | } |
313 | ||
314 | static void nilfs_put_super(struct super_block *sb) | |
315 | { | |
316 | struct nilfs_sb_info *sbi = NILFS_SB(sb); | |
317 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
318 | ||
6cfd0148 CH |
319 | lock_kernel(); |
320 | ||
8c85e125 CH |
321 | if (sb->s_dirt) |
322 | nilfs_write_super(sb); | |
323 | ||
783f6184 RK |
324 | nilfs_detach_segment_constructor(sbi); |
325 | ||
326 | if (!(sb->s_flags & MS_RDONLY)) { | |
327 | down_write(&nilfs->ns_sem); | |
e339ad31 RK |
328 | nilfs->ns_sbp[0]->s_state = cpu_to_le16(nilfs->ns_mount_state); |
329 | nilfs_commit_super(sbi, 1); | |
783f6184 RK |
330 | up_write(&nilfs->ns_sem); |
331 | } | |
332 | ||
333 | nilfs_detach_checkpoint(sbi); | |
334 | put_nilfs(sbi->s_nilfs); | |
335 | sbi->s_super = NULL; | |
336 | sb->s_fs_info = NULL; | |
337 | kfree(sbi); | |
6cfd0148 CH |
338 | |
339 | unlock_kernel(); | |
783f6184 RK |
340 | } |
341 | ||
342 | /** | |
343 | * nilfs_write_super - write super block(s) of NILFS | |
344 | * @sb: super_block | |
345 | * | |
346 | * nilfs_write_super() gets a fs-dependent lock, writes super block(s), and | |
347 | * clears s_dirt. This function is called in the section protected by | |
348 | * lock_super(). | |
349 | * | |
350 | * The s_dirt flag is managed by each filesystem and we protect it by ns_sem | |
351 | * of the struct the_nilfs. Lock order must be as follows: | |
352 | * | |
353 | * 1. lock_super() | |
354 | * 2. down_write(&nilfs->ns_sem) | |
355 | * | |
356 | * Inside NILFS, locking ns_sem is enough to protect s_dirt and the buffer | |
e339ad31 | 357 | * of the super block (nilfs->ns_sbp[]). |
783f6184 RK |
358 | * |
359 | * In most cases, VFS functions call lock_super() before calling these | |
360 | * methods. So we must be careful not to bring on deadlocks when using | |
361 | * lock_super(); see generic_shutdown_super(), write_super(), and so on. | |
362 | * | |
363 | * Note that order of lock_kernel() and lock_super() depends on contexts | |
364 | * of VFS. We should also note that lock_kernel() can be used in its | |
365 | * protective section and only the outermost one has an effect. | |
366 | */ | |
367 | static void nilfs_write_super(struct super_block *sb) | |
368 | { | |
369 | struct nilfs_sb_info *sbi = NILFS_SB(sb); | |
370 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
371 | ||
372 | down_write(&nilfs->ns_sem); | |
e339ad31 RK |
373 | if (!(sb->s_flags & MS_RDONLY)) { |
374 | struct nilfs_super_block **sbp = nilfs->ns_sbp; | |
375 | u64 t = get_seconds(); | |
376 | int dupsb; | |
377 | ||
378 | if (!nilfs_discontinued(nilfs) && t >= nilfs->ns_sbwtime[0] && | |
379 | t < nilfs->ns_sbwtime[0] + NILFS_SB_FREQ) { | |
380 | up_write(&nilfs->ns_sem); | |
381 | return; | |
382 | } | |
383 | dupsb = sbp[1] && t > nilfs->ns_sbwtime[1] + NILFS_ALTSB_FREQ; | |
384 | nilfs_commit_super(sbi, dupsb); | |
385 | } | |
783f6184 RK |
386 | sb->s_dirt = 0; |
387 | up_write(&nilfs->ns_sem); | |
388 | } | |
389 | ||
390 | static int nilfs_sync_fs(struct super_block *sb, int wait) | |
391 | { | |
392 | int err = 0; | |
393 | ||
394 | /* This function is called when super block should be written back */ | |
395 | if (wait) | |
396 | err = nilfs_construct_segment(sb); | |
397 | return err; | |
398 | } | |
399 | ||
400 | int nilfs_attach_checkpoint(struct nilfs_sb_info *sbi, __u64 cno) | |
401 | { | |
402 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
403 | struct nilfs_checkpoint *raw_cp; | |
404 | struct buffer_head *bh_cp; | |
405 | int err; | |
406 | ||
407 | down_write(&nilfs->ns_sem); | |
408 | list_add(&sbi->s_list, &nilfs->ns_supers); | |
409 | up_write(&nilfs->ns_sem); | |
410 | ||
411 | sbi->s_ifile = nilfs_mdt_new( | |
412 | nilfs, sbi->s_super, NILFS_IFILE_INO, NILFS_IFILE_GFP); | |
413 | if (!sbi->s_ifile) | |
414 | return -ENOMEM; | |
415 | ||
416 | err = nilfs_palloc_init_blockgroup(sbi->s_ifile, nilfs->ns_inode_size); | |
417 | if (unlikely(err)) | |
418 | goto failed; | |
419 | ||
420 | err = nilfs_cpfile_get_checkpoint(nilfs->ns_cpfile, cno, 0, &raw_cp, | |
421 | &bh_cp); | |
422 | if (unlikely(err)) { | |
423 | if (err == -ENOENT || err == -EINVAL) { | |
424 | printk(KERN_ERR | |
425 | "NILFS: Invalid checkpoint " | |
426 | "(checkpoint number=%llu)\n", | |
427 | (unsigned long long)cno); | |
428 | err = -EINVAL; | |
429 | } | |
430 | goto failed; | |
431 | } | |
432 | err = nilfs_read_inode_common(sbi->s_ifile, &raw_cp->cp_ifile_inode); | |
433 | if (unlikely(err)) | |
434 | goto failed_bh; | |
435 | atomic_set(&sbi->s_inodes_count, le64_to_cpu(raw_cp->cp_inodes_count)); | |
436 | atomic_set(&sbi->s_blocks_count, le64_to_cpu(raw_cp->cp_blocks_count)); | |
437 | ||
438 | nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, cno, bh_cp); | |
439 | return 0; | |
440 | ||
441 | failed_bh: | |
442 | nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, cno, bh_cp); | |
443 | failed: | |
444 | nilfs_mdt_destroy(sbi->s_ifile); | |
445 | sbi->s_ifile = NULL; | |
446 | ||
447 | down_write(&nilfs->ns_sem); | |
448 | list_del_init(&sbi->s_list); | |
449 | up_write(&nilfs->ns_sem); | |
450 | ||
451 | return err; | |
452 | } | |
453 | ||
454 | void nilfs_detach_checkpoint(struct nilfs_sb_info *sbi) | |
455 | { | |
456 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
457 | ||
458 | nilfs_mdt_clear(sbi->s_ifile); | |
459 | nilfs_mdt_destroy(sbi->s_ifile); | |
460 | sbi->s_ifile = NULL; | |
461 | down_write(&nilfs->ns_sem); | |
462 | list_del_init(&sbi->s_list); | |
463 | up_write(&nilfs->ns_sem); | |
464 | } | |
465 | ||
466 | static int nilfs_mark_recovery_complete(struct nilfs_sb_info *sbi) | |
467 | { | |
468 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
469 | int err = 0; | |
470 | ||
471 | down_write(&nilfs->ns_sem); | |
472 | if (!(nilfs->ns_mount_state & NILFS_VALID_FS)) { | |
473 | nilfs->ns_mount_state |= NILFS_VALID_FS; | |
e339ad31 | 474 | err = nilfs_commit_super(sbi, 1); |
783f6184 RK |
475 | if (likely(!err)) |
476 | printk(KERN_INFO "NILFS: recovery complete.\n"); | |
477 | } | |
478 | up_write(&nilfs->ns_sem); | |
479 | return err; | |
480 | } | |
481 | ||
482 | static int nilfs_statfs(struct dentry *dentry, struct kstatfs *buf) | |
483 | { | |
484 | struct super_block *sb = dentry->d_sb; | |
485 | struct nilfs_sb_info *sbi = NILFS_SB(sb); | |
c306af23 RK |
486 | struct the_nilfs *nilfs = sbi->s_nilfs; |
487 | u64 id = huge_encode_dev(sb->s_bdev->bd_dev); | |
783f6184 RK |
488 | unsigned long long blocks; |
489 | unsigned long overhead; | |
490 | unsigned long nrsvblocks; | |
491 | sector_t nfreeblocks; | |
783f6184 RK |
492 | int err; |
493 | ||
494 | /* | |
495 | * Compute all of the segment blocks | |
496 | * | |
497 | * The blocks before first segment and after last segment | |
498 | * are excluded. | |
499 | */ | |
500 | blocks = nilfs->ns_blocks_per_segment * nilfs->ns_nsegments | |
501 | - nilfs->ns_first_data_block; | |
502 | nrsvblocks = nilfs->ns_nrsvsegs * nilfs->ns_blocks_per_segment; | |
503 | ||
504 | /* | |
505 | * Compute the overhead | |
506 | * | |
507 | * When distributing meta data blocks outside semgent structure, | |
508 | * We must count them as the overhead. | |
509 | */ | |
510 | overhead = 0; | |
511 | ||
512 | err = nilfs_count_free_blocks(nilfs, &nfreeblocks); | |
513 | if (unlikely(err)) | |
514 | return err; | |
515 | ||
516 | buf->f_type = NILFS_SUPER_MAGIC; | |
517 | buf->f_bsize = sb->s_blocksize; | |
518 | buf->f_blocks = blocks - overhead; | |
519 | buf->f_bfree = nfreeblocks; | |
520 | buf->f_bavail = (buf->f_bfree >= nrsvblocks) ? | |
521 | (buf->f_bfree - nrsvblocks) : 0; | |
522 | buf->f_files = atomic_read(&sbi->s_inodes_count); | |
523 | buf->f_ffree = 0; /* nilfs_count_free_inodes(sb); */ | |
524 | buf->f_namelen = NILFS_NAME_LEN; | |
c306af23 RK |
525 | buf->f_fsid.val[0] = (u32)id; |
526 | buf->f_fsid.val[1] = (u32)(id >> 32); | |
527 | ||
783f6184 RK |
528 | return 0; |
529 | } | |
530 | ||
531 | static struct super_operations nilfs_sops = { | |
532 | .alloc_inode = nilfs_alloc_inode, | |
533 | .destroy_inode = nilfs_destroy_inode, | |
534 | .dirty_inode = nilfs_dirty_inode, | |
535 | /* .write_inode = nilfs_write_inode, */ | |
536 | /* .put_inode = nilfs_put_inode, */ | |
537 | /* .drop_inode = nilfs_drop_inode, */ | |
538 | .delete_inode = nilfs_delete_inode, | |
539 | .put_super = nilfs_put_super, | |
540 | .write_super = nilfs_write_super, | |
541 | .sync_fs = nilfs_sync_fs, | |
542 | /* .write_super_lockfs */ | |
543 | /* .unlockfs */ | |
544 | .statfs = nilfs_statfs, | |
545 | .remount_fs = nilfs_remount, | |
546 | .clear_inode = nilfs_clear_inode, | |
547 | /* .umount_begin */ | |
548 | /* .show_options */ | |
549 | }; | |
550 | ||
551 | static struct inode * | |
552 | nilfs_nfs_get_inode(struct super_block *sb, u64 ino, u32 generation) | |
553 | { | |
554 | struct inode *inode; | |
555 | ||
556 | if (ino < NILFS_FIRST_INO(sb) && ino != NILFS_ROOT_INO && | |
557 | ino != NILFS_SKETCH_INO) | |
558 | return ERR_PTR(-ESTALE); | |
559 | ||
560 | inode = nilfs_iget(sb, ino); | |
561 | if (IS_ERR(inode)) | |
562 | return ERR_CAST(inode); | |
563 | if (generation && inode->i_generation != generation) { | |
564 | iput(inode); | |
565 | return ERR_PTR(-ESTALE); | |
566 | } | |
567 | ||
568 | return inode; | |
569 | } | |
570 | ||
571 | static struct dentry * | |
572 | nilfs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len, | |
573 | int fh_type) | |
574 | { | |
575 | return generic_fh_to_dentry(sb, fid, fh_len, fh_type, | |
576 | nilfs_nfs_get_inode); | |
577 | } | |
578 | ||
579 | static struct dentry * | |
580 | nilfs_fh_to_parent(struct super_block *sb, struct fid *fid, int fh_len, | |
581 | int fh_type) | |
582 | { | |
583 | return generic_fh_to_parent(sb, fid, fh_len, fh_type, | |
584 | nilfs_nfs_get_inode); | |
585 | } | |
586 | ||
587 | static struct export_operations nilfs_export_ops = { | |
588 | .fh_to_dentry = nilfs_fh_to_dentry, | |
589 | .fh_to_parent = nilfs_fh_to_parent, | |
590 | .get_parent = nilfs_get_parent, | |
591 | }; | |
592 | ||
593 | enum { | |
594 | Opt_err_cont, Opt_err_panic, Opt_err_ro, | |
595 | Opt_barrier, Opt_snapshot, Opt_order, | |
596 | Opt_err, | |
597 | }; | |
598 | ||
599 | static match_table_t tokens = { | |
600 | {Opt_err_cont, "errors=continue"}, | |
601 | {Opt_err_panic, "errors=panic"}, | |
602 | {Opt_err_ro, "errors=remount-ro"}, | |
603 | {Opt_barrier, "barrier=%s"}, | |
604 | {Opt_snapshot, "cp=%u"}, | |
605 | {Opt_order, "order=%s"}, | |
606 | {Opt_err, NULL} | |
607 | }; | |
608 | ||
609 | static int match_bool(substring_t *s, int *result) | |
610 | { | |
611 | int len = s->to - s->from; | |
612 | ||
613 | if (strncmp(s->from, "on", len) == 0) | |
614 | *result = 1; | |
615 | else if (strncmp(s->from, "off", len) == 0) | |
616 | *result = 0; | |
617 | else | |
618 | return 1; | |
619 | return 0; | |
620 | } | |
621 | ||
622 | static int parse_options(char *options, struct super_block *sb) | |
623 | { | |
624 | struct nilfs_sb_info *sbi = NILFS_SB(sb); | |
625 | char *p; | |
626 | substring_t args[MAX_OPT_ARGS]; | |
627 | int option; | |
628 | ||
629 | if (!options) | |
630 | return 1; | |
631 | ||
632 | while ((p = strsep(&options, ",")) != NULL) { | |
633 | int token; | |
634 | if (!*p) | |
635 | continue; | |
636 | ||
637 | token = match_token(p, tokens, args); | |
638 | switch (token) { | |
639 | case Opt_barrier: | |
640 | if (match_bool(&args[0], &option)) | |
641 | return 0; | |
642 | if (option) | |
643 | nilfs_set_opt(sbi, BARRIER); | |
644 | else | |
645 | nilfs_clear_opt(sbi, BARRIER); | |
646 | break; | |
647 | case Opt_order: | |
648 | if (strcmp(args[0].from, "relaxed") == 0) | |
649 | /* Ordered data semantics */ | |
650 | nilfs_clear_opt(sbi, STRICT_ORDER); | |
651 | else if (strcmp(args[0].from, "strict") == 0) | |
652 | /* Strict in-order semantics */ | |
653 | nilfs_set_opt(sbi, STRICT_ORDER); | |
654 | else | |
655 | return 0; | |
656 | break; | |
657 | case Opt_err_panic: | |
658 | nilfs_write_opt(sbi, ERROR_MODE, ERRORS_PANIC); | |
659 | break; | |
660 | case Opt_err_ro: | |
661 | nilfs_write_opt(sbi, ERROR_MODE, ERRORS_RO); | |
662 | break; | |
663 | case Opt_err_cont: | |
664 | nilfs_write_opt(sbi, ERROR_MODE, ERRORS_CONT); | |
665 | break; | |
666 | case Opt_snapshot: | |
667 | if (match_int(&args[0], &option) || option <= 0) | |
668 | return 0; | |
669 | if (!(sb->s_flags & MS_RDONLY)) | |
670 | return 0; | |
671 | sbi->s_snapshot_cno = option; | |
672 | nilfs_set_opt(sbi, SNAPSHOT); | |
673 | break; | |
674 | default: | |
675 | printk(KERN_ERR | |
676 | "NILFS: Unrecognized mount option \"%s\"\n", p); | |
677 | return 0; | |
678 | } | |
679 | } | |
680 | return 1; | |
681 | } | |
682 | ||
683 | static inline void | |
684 | nilfs_set_default_options(struct nilfs_sb_info *sbi, | |
685 | struct nilfs_super_block *sbp) | |
686 | { | |
687 | sbi->s_mount_opt = | |
688 | NILFS_MOUNT_ERRORS_CONT | NILFS_MOUNT_BARRIER; | |
689 | } | |
690 | ||
691 | static int nilfs_setup_super(struct nilfs_sb_info *sbi) | |
692 | { | |
693 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
e339ad31 | 694 | struct nilfs_super_block *sbp = nilfs->ns_sbp[0]; |
783f6184 RK |
695 | int max_mnt_count = le16_to_cpu(sbp->s_max_mnt_count); |
696 | int mnt_count = le16_to_cpu(sbp->s_mnt_count); | |
697 | ||
698 | /* nilfs->sem must be locked by the caller. */ | |
699 | if (!(nilfs->ns_mount_state & NILFS_VALID_FS)) { | |
700 | printk(KERN_WARNING "NILFS warning: mounting unchecked fs\n"); | |
701 | } else if (nilfs->ns_mount_state & NILFS_ERROR_FS) { | |
702 | printk(KERN_WARNING | |
703 | "NILFS warning: mounting fs with errors\n"); | |
704 | #if 0 | |
705 | } else if (max_mnt_count >= 0 && mnt_count >= max_mnt_count) { | |
706 | printk(KERN_WARNING | |
707 | "NILFS warning: maximal mount count reached\n"); | |
708 | #endif | |
709 | } | |
710 | if (!max_mnt_count) | |
711 | sbp->s_max_mnt_count = cpu_to_le16(NILFS_DFL_MAX_MNT_COUNT); | |
712 | ||
713 | sbp->s_mnt_count = cpu_to_le16(mnt_count + 1); | |
714 | sbp->s_state = cpu_to_le16(le16_to_cpu(sbp->s_state) & ~NILFS_VALID_FS); | |
715 | sbp->s_mtime = cpu_to_le64(get_seconds()); | |
e339ad31 | 716 | return nilfs_commit_super(sbi, 1); |
783f6184 RK |
717 | } |
718 | ||
e339ad31 RK |
719 | struct nilfs_super_block *nilfs_read_super_block(struct super_block *sb, |
720 | u64 pos, int blocksize, | |
721 | struct buffer_head **pbh) | |
783f6184 | 722 | { |
e339ad31 RK |
723 | unsigned long long sb_index = pos; |
724 | unsigned long offset; | |
783f6184 | 725 | |
e339ad31 | 726 | offset = do_div(sb_index, blocksize); |
783f6184 | 727 | *pbh = sb_bread(sb, sb_index); |
e339ad31 | 728 | if (!*pbh) |
783f6184 | 729 | return NULL; |
783f6184 RK |
730 | return (struct nilfs_super_block *)((char *)(*pbh)->b_data + offset); |
731 | } | |
732 | ||
783f6184 RK |
733 | int nilfs_store_magic_and_option(struct super_block *sb, |
734 | struct nilfs_super_block *sbp, | |
735 | char *data) | |
736 | { | |
737 | struct nilfs_sb_info *sbi = NILFS_SB(sb); | |
738 | ||
783f6184 RK |
739 | sb->s_magic = le16_to_cpu(sbp->s_magic); |
740 | ||
741 | /* FS independent flags */ | |
742 | #ifdef NILFS_ATIME_DISABLE | |
743 | sb->s_flags |= MS_NOATIME; | |
744 | #endif | |
745 | ||
783f6184 RK |
746 | nilfs_set_default_options(sbi, sbp); |
747 | ||
748 | sbi->s_resuid = le16_to_cpu(sbp->s_def_resuid); | |
749 | sbi->s_resgid = le16_to_cpu(sbp->s_def_resgid); | |
750 | sbi->s_interval = le32_to_cpu(sbp->s_c_interval); | |
751 | sbi->s_watermark = le32_to_cpu(sbp->s_c_block_max); | |
752 | ||
e339ad31 | 753 | return !parse_options(data, sb) ? -EINVAL : 0 ; |
783f6184 RK |
754 | } |
755 | ||
756 | /** | |
757 | * nilfs_fill_super() - initialize a super block instance | |
758 | * @sb: super_block | |
759 | * @data: mount options | |
760 | * @silent: silent mode flag | |
761 | * @nilfs: the_nilfs struct | |
762 | * | |
763 | * This function is called exclusively by bd_mount_mutex. | |
764 | * So, the recovery process is protected from other simultaneous mounts. | |
765 | */ | |
766 | static int | |
767 | nilfs_fill_super(struct super_block *sb, void *data, int silent, | |
768 | struct the_nilfs *nilfs) | |
769 | { | |
770 | struct nilfs_sb_info *sbi; | |
771 | struct inode *root; | |
772 | __u64 cno; | |
773 | int err; | |
774 | ||
775 | sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); | |
776 | if (!sbi) | |
777 | return -ENOMEM; | |
778 | ||
779 | sb->s_fs_info = sbi; | |
780 | ||
781 | get_nilfs(nilfs); | |
782 | sbi->s_nilfs = nilfs; | |
783 | sbi->s_super = sb; | |
784 | ||
785 | err = init_nilfs(nilfs, sbi, (char *)data); | |
786 | if (err) | |
787 | goto failed_sbi; | |
788 | ||
789 | spin_lock_init(&sbi->s_inode_lock); | |
790 | INIT_LIST_HEAD(&sbi->s_dirty_files); | |
791 | INIT_LIST_HEAD(&sbi->s_list); | |
792 | ||
793 | /* | |
794 | * Following initialization is overlapped because | |
795 | * nilfs_sb_info structure has been cleared at the beginning. | |
796 | * But we reserve them to keep our interest and make ready | |
797 | * for the future change. | |
798 | */ | |
799 | get_random_bytes(&sbi->s_next_generation, | |
800 | sizeof(sbi->s_next_generation)); | |
801 | spin_lock_init(&sbi->s_next_gen_lock); | |
802 | ||
803 | sb->s_op = &nilfs_sops; | |
804 | sb->s_export_op = &nilfs_export_ops; | |
805 | sb->s_root = NULL; | |
61239230 | 806 | sb->s_time_gran = 1; |
783f6184 RK |
807 | |
808 | if (!nilfs_loaded(nilfs)) { | |
809 | err = load_nilfs(nilfs, sbi); | |
810 | if (err) | |
811 | goto failed_sbi; | |
812 | } | |
813 | cno = nilfs_last_cno(nilfs); | |
814 | ||
815 | if (sb->s_flags & MS_RDONLY) { | |
816 | if (nilfs_test_opt(sbi, SNAPSHOT)) { | |
1f5abe7e RK |
817 | err = nilfs_cpfile_is_snapshot(nilfs->ns_cpfile, |
818 | sbi->s_snapshot_cno); | |
819 | if (err < 0) | |
820 | goto failed_sbi; | |
821 | if (!err) { | |
783f6184 RK |
822 | printk(KERN_ERR |
823 | "NILFS: The specified checkpoint is " | |
824 | "not a snapshot " | |
825 | "(checkpoint number=%llu).\n", | |
826 | (unsigned long long)sbi->s_snapshot_cno); | |
827 | err = -EINVAL; | |
828 | goto failed_sbi; | |
829 | } | |
830 | cno = sbi->s_snapshot_cno; | |
831 | } else | |
832 | /* Read-only mount */ | |
833 | sbi->s_snapshot_cno = cno; | |
834 | } | |
835 | ||
836 | err = nilfs_attach_checkpoint(sbi, cno); | |
837 | if (err) { | |
838 | printk(KERN_ERR "NILFS: error loading a checkpoint" | |
839 | " (checkpoint number=%llu).\n", (unsigned long long)cno); | |
840 | goto failed_sbi; | |
841 | } | |
842 | ||
843 | if (!(sb->s_flags & MS_RDONLY)) { | |
cece5520 | 844 | err = nilfs_attach_segment_constructor(sbi); |
783f6184 RK |
845 | if (err) |
846 | goto failed_checkpoint; | |
847 | } | |
848 | ||
849 | root = nilfs_iget(sb, NILFS_ROOT_INO); | |
850 | if (IS_ERR(root)) { | |
851 | printk(KERN_ERR "NILFS: get root inode failed\n"); | |
852 | err = PTR_ERR(root); | |
853 | goto failed_segctor; | |
854 | } | |
855 | if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { | |
856 | iput(root); | |
857 | printk(KERN_ERR "NILFS: corrupt root inode.\n"); | |
858 | err = -EINVAL; | |
859 | goto failed_segctor; | |
860 | } | |
861 | sb->s_root = d_alloc_root(root); | |
862 | if (!sb->s_root) { | |
863 | iput(root); | |
864 | printk(KERN_ERR "NILFS: get root dentry failed\n"); | |
865 | err = -ENOMEM; | |
866 | goto failed_segctor; | |
867 | } | |
868 | ||
869 | if (!(sb->s_flags & MS_RDONLY)) { | |
870 | down_write(&nilfs->ns_sem); | |
871 | nilfs_setup_super(sbi); | |
872 | up_write(&nilfs->ns_sem); | |
873 | } | |
874 | ||
875 | err = nilfs_mark_recovery_complete(sbi); | |
876 | if (unlikely(err)) { | |
877 | printk(KERN_ERR "NILFS: recovery failed.\n"); | |
878 | goto failed_root; | |
879 | } | |
880 | ||
881 | return 0; | |
882 | ||
883 | failed_root: | |
884 | dput(sb->s_root); | |
885 | sb->s_root = NULL; | |
886 | ||
887 | failed_segctor: | |
888 | nilfs_detach_segment_constructor(sbi); | |
889 | ||
890 | failed_checkpoint: | |
891 | nilfs_detach_checkpoint(sbi); | |
892 | ||
893 | failed_sbi: | |
894 | put_nilfs(nilfs); | |
895 | sb->s_fs_info = NULL; | |
896 | kfree(sbi); | |
897 | return err; | |
898 | } | |
899 | ||
900 | static int nilfs_remount(struct super_block *sb, int *flags, char *data) | |
901 | { | |
902 | struct nilfs_sb_info *sbi = NILFS_SB(sb); | |
903 | struct nilfs_super_block *sbp; | |
904 | struct the_nilfs *nilfs = sbi->s_nilfs; | |
905 | unsigned long old_sb_flags; | |
906 | struct nilfs_mount_options old_opts; | |
907 | int err; | |
908 | ||
909 | old_sb_flags = sb->s_flags; | |
910 | old_opts.mount_opt = sbi->s_mount_opt; | |
911 | old_opts.snapshot_cno = sbi->s_snapshot_cno; | |
912 | ||
913 | if (!parse_options(data, sb)) { | |
914 | err = -EINVAL; | |
915 | goto restore_opts; | |
916 | } | |
917 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL); | |
918 | ||
919 | if ((*flags & MS_RDONLY) && | |
920 | sbi->s_snapshot_cno != old_opts.snapshot_cno) { | |
921 | printk(KERN_WARNING "NILFS (device %s): couldn't " | |
922 | "remount to a different snapshot. \n", | |
923 | sb->s_id); | |
924 | err = -EINVAL; | |
925 | goto restore_opts; | |
926 | } | |
927 | ||
928 | if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) | |
929 | goto out; | |
930 | if (*flags & MS_RDONLY) { | |
931 | /* Shutting down the segment constructor */ | |
932 | nilfs_detach_segment_constructor(sbi); | |
933 | sb->s_flags |= MS_RDONLY; | |
934 | ||
935 | sbi->s_snapshot_cno = nilfs_last_cno(nilfs); | |
936 | /* nilfs_set_opt(sbi, SNAPSHOT); */ | |
937 | ||
938 | /* | |
939 | * Remounting a valid RW partition RDONLY, so set | |
940 | * the RDONLY flag and then mark the partition as valid again. | |
941 | */ | |
942 | down_write(&nilfs->ns_sem); | |
e339ad31 | 943 | sbp = nilfs->ns_sbp[0]; |
783f6184 RK |
944 | if (!(sbp->s_state & le16_to_cpu(NILFS_VALID_FS)) && |
945 | (nilfs->ns_mount_state & NILFS_VALID_FS)) | |
946 | sbp->s_state = cpu_to_le16(nilfs->ns_mount_state); | |
947 | sbp->s_mtime = cpu_to_le64(get_seconds()); | |
e339ad31 | 948 | nilfs_commit_super(sbi, 1); |
783f6184 RK |
949 | up_write(&nilfs->ns_sem); |
950 | } else { | |
951 | /* | |
952 | * Mounting a RDONLY partition read-write, so reread and | |
953 | * store the current valid flag. (It may have been changed | |
954 | * by fsck since we originally mounted the partition.) | |
955 | */ | |
956 | down(&sb->s_bdev->bd_mount_sem); | |
957 | /* Check existing RW-mount */ | |
958 | if (test_exclusive_mount(sb->s_type, sb->s_bdev, 0)) { | |
959 | printk(KERN_WARNING "NILFS (device %s): couldn't " | |
960 | "remount because a RW-mount exists.\n", | |
961 | sb->s_id); | |
962 | err = -EBUSY; | |
963 | goto rw_remount_failed; | |
964 | } | |
965 | if (sbi->s_snapshot_cno != nilfs_last_cno(nilfs)) { | |
966 | printk(KERN_WARNING "NILFS (device %s): couldn't " | |
967 | "remount because the current RO-mount is not " | |
968 | "the latest one.\n", | |
969 | sb->s_id); | |
970 | err = -EINVAL; | |
971 | goto rw_remount_failed; | |
972 | } | |
973 | sb->s_flags &= ~MS_RDONLY; | |
974 | nilfs_clear_opt(sbi, SNAPSHOT); | |
975 | sbi->s_snapshot_cno = 0; | |
976 | ||
cece5520 | 977 | err = nilfs_attach_segment_constructor(sbi); |
783f6184 RK |
978 | if (err) |
979 | goto rw_remount_failed; | |
980 | ||
981 | down_write(&nilfs->ns_sem); | |
982 | nilfs_setup_super(sbi); | |
983 | up_write(&nilfs->ns_sem); | |
984 | ||
985 | up(&sb->s_bdev->bd_mount_sem); | |
986 | } | |
987 | out: | |
988 | return 0; | |
989 | ||
990 | rw_remount_failed: | |
991 | up(&sb->s_bdev->bd_mount_sem); | |
992 | restore_opts: | |
993 | sb->s_flags = old_sb_flags; | |
994 | sbi->s_mount_opt = old_opts.mount_opt; | |
995 | sbi->s_snapshot_cno = old_opts.snapshot_cno; | |
996 | return err; | |
997 | } | |
998 | ||
999 | struct nilfs_super_data { | |
1000 | struct block_device *bdev; | |
1001 | __u64 cno; | |
1002 | int flags; | |
1003 | }; | |
1004 | ||
1005 | /** | |
1006 | * nilfs_identify - pre-read mount options needed to identify mount instance | |
1007 | * @data: mount options | |
1008 | * @sd: nilfs_super_data | |
1009 | */ | |
1010 | static int nilfs_identify(char *data, struct nilfs_super_data *sd) | |
1011 | { | |
1012 | char *p, *options = data; | |
1013 | substring_t args[MAX_OPT_ARGS]; | |
1014 | int option, token; | |
1015 | int ret = 0; | |
1016 | ||
1017 | do { | |
1018 | p = strsep(&options, ","); | |
1019 | if (p != NULL && *p) { | |
1020 | token = match_token(p, tokens, args); | |
1021 | if (token == Opt_snapshot) { | |
1022 | if (!(sd->flags & MS_RDONLY)) | |
1023 | ret++; | |
1024 | else { | |
1025 | ret = match_int(&args[0], &option); | |
1026 | if (!ret) { | |
1027 | if (option > 0) | |
1028 | sd->cno = option; | |
1029 | else | |
1030 | ret++; | |
1031 | } | |
1032 | } | |
1033 | } | |
1034 | if (ret) | |
1035 | printk(KERN_ERR | |
1036 | "NILFS: invalid mount option: %s\n", p); | |
1037 | } | |
1038 | if (!options) | |
1039 | break; | |
1040 | BUG_ON(options == data); | |
1041 | *(options - 1) = ','; | |
1042 | } while (!ret); | |
1043 | return ret; | |
1044 | } | |
1045 | ||
1046 | static int nilfs_set_bdev_super(struct super_block *s, void *data) | |
1047 | { | |
1048 | struct nilfs_super_data *sd = data; | |
1049 | ||
1050 | s->s_bdev = sd->bdev; | |
1051 | s->s_dev = s->s_bdev->bd_dev; | |
1052 | return 0; | |
1053 | } | |
1054 | ||
1055 | static int nilfs_test_bdev_super(struct super_block *s, void *data) | |
1056 | { | |
1057 | struct nilfs_super_data *sd = data; | |
1058 | ||
1059 | return s->s_bdev == sd->bdev; | |
1060 | } | |
1061 | ||
1062 | static int nilfs_test_bdev_super2(struct super_block *s, void *data) | |
1063 | { | |
1064 | struct nilfs_super_data *sd = data; | |
1065 | int ret; | |
1066 | ||
1067 | if (s->s_bdev != sd->bdev) | |
1068 | return 0; | |
1069 | ||
1070 | if (!((s->s_flags | sd->flags) & MS_RDONLY)) | |
1071 | return 1; /* Reuse an old R/W-mode super_block */ | |
1072 | ||
1073 | if (s->s_flags & sd->flags & MS_RDONLY) { | |
1074 | if (down_read_trylock(&s->s_umount)) { | |
1075 | ret = s->s_root && | |
1076 | (sd->cno == NILFS_SB(s)->s_snapshot_cno); | |
1077 | up_read(&s->s_umount); | |
1078 | /* | |
1079 | * This path is locked with sb_lock by sget(). | |
1080 | * So, drop_super() causes deadlock. | |
1081 | */ | |
1082 | return ret; | |
1083 | } | |
1084 | } | |
1085 | return 0; | |
1086 | } | |
1087 | ||
1088 | static int | |
1089 | nilfs_get_sb(struct file_system_type *fs_type, int flags, | |
1090 | const char *dev_name, void *data, struct vfsmount *mnt) | |
1091 | { | |
1092 | struct nilfs_super_data sd; | |
1093 | struct super_block *s, *s2; | |
1094 | struct the_nilfs *nilfs = NULL; | |
1095 | int err, need_to_close = 1; | |
1096 | ||
1097 | sd.bdev = open_bdev_exclusive(dev_name, flags, fs_type); | |
1098 | if (IS_ERR(sd.bdev)) | |
1099 | return PTR_ERR(sd.bdev); | |
1100 | ||
1101 | /* | |
1102 | * To get mount instance using sget() vfs-routine, NILFS needs | |
1103 | * much more information than normal filesystems to identify mount | |
1104 | * instance. For snapshot mounts, not only a mount type (ro-mount | |
1105 | * or rw-mount) but also a checkpoint number is required. | |
1106 | * The results are passed in sget() using nilfs_super_data. | |
1107 | */ | |
1108 | sd.cno = 0; | |
1109 | sd.flags = flags; | |
1110 | if (nilfs_identify((char *)data, &sd)) { | |
1111 | err = -EINVAL; | |
1112 | goto failed; | |
1113 | } | |
1114 | ||
1115 | /* | |
1116 | * once the super is inserted into the list by sget, s_umount | |
1117 | * will protect the lockfs code from trying to start a snapshot | |
1118 | * while we are mounting | |
1119 | */ | |
1120 | down(&sd.bdev->bd_mount_sem); | |
1121 | if (!sd.cno && | |
1122 | (err = test_exclusive_mount(fs_type, sd.bdev, flags ^ MS_RDONLY))) { | |
1123 | err = (err < 0) ? : -EBUSY; | |
1124 | goto failed_unlock; | |
1125 | } | |
1126 | ||
1127 | /* | |
1128 | * Phase-1: search any existent instance and get the_nilfs | |
1129 | */ | |
1130 | s = sget(fs_type, nilfs_test_bdev_super, nilfs_set_bdev_super, &sd); | |
1131 | if (IS_ERR(s)) | |
1132 | goto error_s; | |
1133 | ||
1134 | if (!s->s_root) { | |
1135 | err = -ENOMEM; | |
1136 | nilfs = alloc_nilfs(sd.bdev); | |
1137 | if (!nilfs) | |
1138 | goto cancel_new; | |
1139 | } else { | |
1140 | struct nilfs_sb_info *sbi = NILFS_SB(s); | |
1141 | ||
783f6184 RK |
1142 | /* |
1143 | * s_umount protects super_block from unmount process; | |
1144 | * It covers pointers of nilfs_sb_info and the_nilfs. | |
1145 | */ | |
1146 | nilfs = sbi->s_nilfs; | |
1147 | get_nilfs(nilfs); | |
1148 | up_write(&s->s_umount); | |
1149 | ||
1150 | /* | |
1151 | * Phase-2: search specified snapshot or R/W mode super_block | |
1152 | */ | |
1153 | if (!sd.cno) | |
1154 | /* trying to get the latest checkpoint. */ | |
1155 | sd.cno = nilfs_last_cno(nilfs); | |
1156 | ||
1157 | s2 = sget(fs_type, nilfs_test_bdev_super2, | |
1158 | nilfs_set_bdev_super, &sd); | |
1159 | deactivate_super(s); | |
1160 | /* | |
1161 | * Although deactivate_super() invokes close_bdev_exclusive() at | |
1162 | * kill_block_super(). Here, s is an existent mount; we need | |
1163 | * one more close_bdev_exclusive() call. | |
1164 | */ | |
1165 | s = s2; | |
1166 | if (IS_ERR(s)) | |
1167 | goto error_s; | |
1168 | } | |
1169 | ||
1170 | if (!s->s_root) { | |
1171 | char b[BDEVNAME_SIZE]; | |
1172 | ||
1173 | s->s_flags = flags; | |
1174 | strlcpy(s->s_id, bdevname(sd.bdev, b), sizeof(s->s_id)); | |
1175 | sb_set_blocksize(s, block_size(sd.bdev)); | |
1176 | ||
1177 | err = nilfs_fill_super(s, data, flags & MS_VERBOSE, nilfs); | |
1178 | if (err) | |
1179 | goto cancel_new; | |
1180 | ||
1181 | s->s_flags |= MS_ACTIVE; | |
1182 | need_to_close = 0; | |
1183 | } else if (!(s->s_flags & MS_RDONLY)) { | |
1184 | err = -EBUSY; | |
1185 | } | |
1186 | ||
1187 | up(&sd.bdev->bd_mount_sem); | |
1188 | put_nilfs(nilfs); | |
1189 | if (need_to_close) | |
1190 | close_bdev_exclusive(sd.bdev, flags); | |
1191 | simple_set_mnt(mnt, s); | |
1192 | return 0; | |
1193 | ||
1194 | error_s: | |
1195 | up(&sd.bdev->bd_mount_sem); | |
1196 | if (nilfs) | |
1197 | put_nilfs(nilfs); | |
1198 | close_bdev_exclusive(sd.bdev, flags); | |
1199 | return PTR_ERR(s); | |
1200 | ||
1201 | failed_unlock: | |
1202 | up(&sd.bdev->bd_mount_sem); | |
1203 | failed: | |
1204 | close_bdev_exclusive(sd.bdev, flags); | |
1205 | ||
1206 | return err; | |
1207 | ||
1208 | cancel_new: | |
1209 | /* Abandoning the newly allocated superblock */ | |
1210 | up(&sd.bdev->bd_mount_sem); | |
1211 | if (nilfs) | |
1212 | put_nilfs(nilfs); | |
1213 | up_write(&s->s_umount); | |
1214 | deactivate_super(s); | |
1215 | /* | |
1216 | * deactivate_super() invokes close_bdev_exclusive(). | |
1217 | * We must finish all post-cleaning before this call; | |
1218 | * put_nilfs() and unlocking bd_mount_sem need the block device. | |
1219 | */ | |
1220 | return err; | |
1221 | } | |
1222 | ||
1223 | static int nilfs_test_bdev_super3(struct super_block *s, void *data) | |
1224 | { | |
1225 | struct nilfs_super_data *sd = data; | |
1226 | int ret; | |
1227 | ||
1228 | if (s->s_bdev != sd->bdev) | |
1229 | return 0; | |
1230 | if (down_read_trylock(&s->s_umount)) { | |
1231 | ret = (s->s_flags & MS_RDONLY) && s->s_root && | |
1232 | nilfs_test_opt(NILFS_SB(s), SNAPSHOT); | |
1233 | up_read(&s->s_umount); | |
1234 | if (ret) | |
1235 | return 0; /* ignore snapshot mounts */ | |
1236 | } | |
1237 | return !((sd->flags ^ s->s_flags) & MS_RDONLY); | |
1238 | } | |
1239 | ||
1240 | static int __false_bdev_super(struct super_block *s, void *data) | |
1241 | { | |
1242 | #if 0 /* XXX: workaround for lock debug. This is not good idea */ | |
1243 | up_write(&s->s_umount); | |
1244 | #endif | |
1245 | return -EFAULT; | |
1246 | } | |
1247 | ||
1248 | /** | |
1249 | * test_exclusive_mount - check whether an exclusive RW/RO mount exists or not. | |
1250 | * fs_type: filesystem type | |
1251 | * bdev: block device | |
1252 | * flag: 0 (check rw-mount) or MS_RDONLY (check ro-mount) | |
1253 | * res: pointer to an integer to store result | |
1254 | * | |
1255 | * This function must be called within a section protected by bd_mount_mutex. | |
1256 | */ | |
1257 | static int test_exclusive_mount(struct file_system_type *fs_type, | |
1258 | struct block_device *bdev, int flags) | |
1259 | { | |
1260 | struct super_block *s; | |
1261 | struct nilfs_super_data sd = { .flags = flags, .bdev = bdev }; | |
1262 | ||
1263 | s = sget(fs_type, nilfs_test_bdev_super3, __false_bdev_super, &sd); | |
1264 | if (IS_ERR(s)) { | |
1265 | if (PTR_ERR(s) != -EFAULT) | |
1266 | return PTR_ERR(s); | |
1267 | return 0; /* Not found */ | |
1268 | } | |
1269 | up_write(&s->s_umount); | |
1270 | deactivate_super(s); | |
1271 | return 1; /* Found */ | |
1272 | } | |
1273 | ||
1274 | struct file_system_type nilfs_fs_type = { | |
1275 | .owner = THIS_MODULE, | |
1276 | .name = "nilfs2", | |
1277 | .get_sb = nilfs_get_sb, | |
1278 | .kill_sb = kill_block_super, | |
1279 | .fs_flags = FS_REQUIRES_DEV, | |
1280 | }; | |
1281 | ||
1282 | static int __init init_nilfs_fs(void) | |
1283 | { | |
1284 | int err; | |
1285 | ||
1286 | err = nilfs_init_inode_cache(); | |
1287 | if (err) | |
1288 | goto failed; | |
1289 | ||
1290 | err = nilfs_init_transaction_cache(); | |
1291 | if (err) | |
1292 | goto failed_inode_cache; | |
1293 | ||
1294 | err = nilfs_init_segbuf_cache(); | |
1295 | if (err) | |
1296 | goto failed_transaction_cache; | |
1297 | ||
1298 | err = nilfs_btree_path_cache_init(); | |
1299 | if (err) | |
1300 | goto failed_segbuf_cache; | |
1301 | ||
1302 | err = register_filesystem(&nilfs_fs_type); | |
1303 | if (err) | |
1304 | goto failed_btree_path_cache; | |
1305 | ||
1306 | return 0; | |
1307 | ||
1308 | failed_btree_path_cache: | |
1309 | nilfs_btree_path_cache_destroy(); | |
1310 | ||
1311 | failed_segbuf_cache: | |
1312 | nilfs_destroy_segbuf_cache(); | |
1313 | ||
1314 | failed_transaction_cache: | |
1315 | nilfs_destroy_transaction_cache(); | |
1316 | ||
1317 | failed_inode_cache: | |
1318 | nilfs_destroy_inode_cache(); | |
1319 | ||
1320 | failed: | |
1321 | return err; | |
1322 | } | |
1323 | ||
1324 | static void __exit exit_nilfs_fs(void) | |
1325 | { | |
1326 | nilfs_destroy_segbuf_cache(); | |
1327 | nilfs_destroy_transaction_cache(); | |
1328 | nilfs_destroy_inode_cache(); | |
1329 | nilfs_btree_path_cache_destroy(); | |
1330 | unregister_filesystem(&nilfs_fs_type); | |
1331 | } | |
1332 | ||
1333 | module_init(init_nilfs_fs) | |
1334 | module_exit(exit_nilfs_fs) |