]>
Commit | Line | Data |
---|---|---|
8a9d2191 RK |
1 | /* |
2 | * the_nilfs.c - the_nilfs shared structure. | |
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 | ||
24 | #include <linux/buffer_head.h> | |
25 | #include <linux/slab.h> | |
26 | #include <linux/blkdev.h> | |
27 | #include <linux/backing-dev.h> | |
28 | #include "nilfs.h" | |
29 | #include "segment.h" | |
30 | #include "alloc.h" | |
31 | #include "cpfile.h" | |
32 | #include "sufile.h" | |
33 | #include "dat.h" | |
34 | #include "seglist.h" | |
35 | #include "segbuf.h" | |
36 | ||
37 | void nilfs_set_last_segment(struct the_nilfs *nilfs, | |
38 | sector_t start_blocknr, u64 seq, __u64 cno) | |
39 | { | |
40 | spin_lock(&nilfs->ns_last_segment_lock); | |
41 | nilfs->ns_last_pseg = start_blocknr; | |
42 | nilfs->ns_last_seq = seq; | |
43 | nilfs->ns_last_cno = cno; | |
44 | spin_unlock(&nilfs->ns_last_segment_lock); | |
45 | } | |
46 | ||
47 | /** | |
48 | * alloc_nilfs - allocate the_nilfs structure | |
49 | * @bdev: block device to which the_nilfs is related | |
50 | * | |
51 | * alloc_nilfs() allocates memory for the_nilfs and | |
52 | * initializes its reference count and locks. | |
53 | * | |
54 | * Return Value: On success, pointer to the_nilfs is returned. | |
55 | * On error, NULL is returned. | |
56 | */ | |
57 | struct the_nilfs *alloc_nilfs(struct block_device *bdev) | |
58 | { | |
59 | struct the_nilfs *nilfs; | |
60 | ||
61 | nilfs = kzalloc(sizeof(*nilfs), GFP_KERNEL); | |
62 | if (!nilfs) | |
63 | return NULL; | |
64 | ||
65 | nilfs->ns_bdev = bdev; | |
66 | atomic_set(&nilfs->ns_count, 1); | |
67 | atomic_set(&nilfs->ns_writer_refcount, -1); | |
68 | atomic_set(&nilfs->ns_ndirtyblks, 0); | |
69 | init_rwsem(&nilfs->ns_sem); | |
70 | mutex_init(&nilfs->ns_writer_mutex); | |
71 | INIT_LIST_HEAD(&nilfs->ns_supers); | |
72 | spin_lock_init(&nilfs->ns_last_segment_lock); | |
73 | nilfs->ns_gc_inodes_h = NULL; | |
8a9d2191 | 74 | init_rwsem(&nilfs->ns_segctor_sem); |
8a9d2191 RK |
75 | |
76 | return nilfs; | |
77 | } | |
78 | ||
79 | /** | |
80 | * put_nilfs - release a reference to the_nilfs | |
81 | * @nilfs: the_nilfs structure to be released | |
82 | * | |
83 | * put_nilfs() decrements a reference counter of the_nilfs. | |
84 | * If the reference count reaches zero, the_nilfs is freed. | |
85 | */ | |
86 | void put_nilfs(struct the_nilfs *nilfs) | |
87 | { | |
88 | if (!atomic_dec_and_test(&nilfs->ns_count)) | |
89 | return; | |
90 | /* | |
91 | * Increment of ns_count never occur below because the caller | |
92 | * of get_nilfs() holds at least one reference to the_nilfs. | |
93 | * Thus its exclusion control is not required here. | |
94 | */ | |
95 | might_sleep(); | |
96 | if (nilfs_loaded(nilfs)) { | |
8a9d2191 RK |
97 | nilfs_mdt_clear(nilfs->ns_sufile); |
98 | nilfs_mdt_destroy(nilfs->ns_sufile); | |
99 | nilfs_mdt_clear(nilfs->ns_cpfile); | |
100 | nilfs_mdt_destroy(nilfs->ns_cpfile); | |
101 | nilfs_mdt_clear(nilfs->ns_dat); | |
102 | nilfs_mdt_destroy(nilfs->ns_dat); | |
103 | /* XXX: how and when to clear nilfs->ns_gc_dat? */ | |
104 | nilfs_mdt_destroy(nilfs->ns_gc_dat); | |
105 | } | |
106 | if (nilfs_init(nilfs)) { | |
107 | nilfs_destroy_gccache(nilfs); | |
108 | brelse(nilfs->ns_sbh); | |
109 | } | |
110 | kfree(nilfs); | |
111 | } | |
112 | ||
113 | static int nilfs_load_super_root(struct the_nilfs *nilfs, | |
114 | struct nilfs_sb_info *sbi, sector_t sr_block) | |
115 | { | |
116 | struct buffer_head *bh_sr; | |
117 | struct nilfs_super_root *raw_sr; | |
118 | unsigned dat_entry_size, segment_usage_size, checkpoint_size; | |
119 | unsigned inode_size; | |
120 | int err; | |
121 | ||
122 | err = nilfs_read_super_root_block(sbi->s_super, sr_block, &bh_sr, 1); | |
123 | if (unlikely(err)) | |
124 | return err; | |
125 | ||
126 | down_read(&nilfs->ns_sem); | |
127 | dat_entry_size = le16_to_cpu(nilfs->ns_sbp->s_dat_entry_size); | |
128 | checkpoint_size = le16_to_cpu(nilfs->ns_sbp->s_checkpoint_size); | |
129 | segment_usage_size = le16_to_cpu(nilfs->ns_sbp->s_segment_usage_size); | |
130 | up_read(&nilfs->ns_sem); | |
131 | ||
132 | inode_size = nilfs->ns_inode_size; | |
133 | ||
134 | err = -ENOMEM; | |
135 | nilfs->ns_dat = nilfs_mdt_new( | |
136 | nilfs, NULL, NILFS_DAT_INO, NILFS_DAT_GFP); | |
137 | if (unlikely(!nilfs->ns_dat)) | |
138 | goto failed; | |
139 | ||
140 | nilfs->ns_gc_dat = nilfs_mdt_new( | |
141 | nilfs, NULL, NILFS_DAT_INO, NILFS_DAT_GFP); | |
142 | if (unlikely(!nilfs->ns_gc_dat)) | |
143 | goto failed_dat; | |
144 | ||
145 | nilfs->ns_cpfile = nilfs_mdt_new( | |
146 | nilfs, NULL, NILFS_CPFILE_INO, NILFS_CPFILE_GFP); | |
147 | if (unlikely(!nilfs->ns_cpfile)) | |
148 | goto failed_gc_dat; | |
149 | ||
150 | nilfs->ns_sufile = nilfs_mdt_new( | |
151 | nilfs, NULL, NILFS_SUFILE_INO, NILFS_SUFILE_GFP); | |
152 | if (unlikely(!nilfs->ns_sufile)) | |
153 | goto failed_cpfile; | |
154 | ||
155 | err = nilfs_palloc_init_blockgroup(nilfs->ns_dat, dat_entry_size); | |
156 | if (unlikely(err)) | |
157 | goto failed_sufile; | |
158 | ||
159 | err = nilfs_palloc_init_blockgroup(nilfs->ns_gc_dat, dat_entry_size); | |
160 | if (unlikely(err)) | |
161 | goto failed_sufile; | |
162 | ||
163 | nilfs_mdt_set_shadow(nilfs->ns_dat, nilfs->ns_gc_dat); | |
164 | nilfs_mdt_set_entry_size(nilfs->ns_cpfile, checkpoint_size, | |
165 | sizeof(struct nilfs_cpfile_header)); | |
166 | nilfs_mdt_set_entry_size(nilfs->ns_sufile, segment_usage_size, | |
167 | sizeof(struct nilfs_sufile_header)); | |
168 | ||
169 | err = nilfs_mdt_read_inode_direct( | |
170 | nilfs->ns_dat, bh_sr, NILFS_SR_DAT_OFFSET(inode_size)); | |
171 | if (unlikely(err)) | |
172 | goto failed_sufile; | |
173 | ||
174 | err = nilfs_mdt_read_inode_direct( | |
175 | nilfs->ns_cpfile, bh_sr, NILFS_SR_CPFILE_OFFSET(inode_size)); | |
176 | if (unlikely(err)) | |
177 | goto failed_sufile; | |
178 | ||
179 | err = nilfs_mdt_read_inode_direct( | |
180 | nilfs->ns_sufile, bh_sr, NILFS_SR_SUFILE_OFFSET(inode_size)); | |
181 | if (unlikely(err)) | |
182 | goto failed_sufile; | |
183 | ||
184 | raw_sr = (struct nilfs_super_root *)bh_sr->b_data; | |
185 | nilfs->ns_nongc_ctime = le64_to_cpu(raw_sr->sr_nongc_ctime); | |
186 | ||
187 | failed: | |
188 | brelse(bh_sr); | |
189 | return err; | |
190 | ||
191 | failed_sufile: | |
192 | nilfs_mdt_destroy(nilfs->ns_sufile); | |
193 | ||
194 | failed_cpfile: | |
195 | nilfs_mdt_destroy(nilfs->ns_cpfile); | |
196 | ||
197 | failed_gc_dat: | |
198 | nilfs_mdt_destroy(nilfs->ns_gc_dat); | |
199 | ||
200 | failed_dat: | |
201 | nilfs_mdt_destroy(nilfs->ns_dat); | |
202 | goto failed; | |
203 | } | |
204 | ||
205 | static void nilfs_init_recovery_info(struct nilfs_recovery_info *ri) | |
206 | { | |
207 | memset(ri, 0, sizeof(*ri)); | |
208 | INIT_LIST_HEAD(&ri->ri_used_segments); | |
209 | } | |
210 | ||
211 | static void nilfs_clear_recovery_info(struct nilfs_recovery_info *ri) | |
212 | { | |
213 | nilfs_dispose_segment_list(&ri->ri_used_segments); | |
214 | } | |
215 | ||
216 | /** | |
217 | * load_nilfs - load and recover the nilfs | |
218 | * @nilfs: the_nilfs structure to be released | |
219 | * @sbi: nilfs_sb_info used to recover past segment | |
220 | * | |
221 | * load_nilfs() searches and load the latest super root, | |
222 | * attaches the last segment, and does recovery if needed. | |
223 | * The caller must call this exclusively for simultaneous mounts. | |
224 | */ | |
225 | int load_nilfs(struct the_nilfs *nilfs, struct nilfs_sb_info *sbi) | |
226 | { | |
227 | struct nilfs_recovery_info ri; | |
228 | unsigned int s_flags = sbi->s_super->s_flags; | |
229 | int really_read_only = bdev_read_only(nilfs->ns_bdev); | |
230 | unsigned valid_fs; | |
231 | int err = 0; | |
232 | ||
233 | nilfs_init_recovery_info(&ri); | |
234 | ||
235 | down_write(&nilfs->ns_sem); | |
236 | valid_fs = (nilfs->ns_mount_state & NILFS_VALID_FS); | |
237 | up_write(&nilfs->ns_sem); | |
238 | ||
239 | if (!valid_fs && (s_flags & MS_RDONLY)) { | |
240 | printk(KERN_INFO "NILFS: INFO: recovery " | |
241 | "required for readonly filesystem.\n"); | |
242 | if (really_read_only) { | |
243 | printk(KERN_ERR "NILFS: write access " | |
244 | "unavailable, cannot proceed.\n"); | |
245 | err = -EROFS; | |
246 | goto failed; | |
247 | } | |
248 | printk(KERN_INFO "NILFS: write access will " | |
249 | "be enabled during recovery.\n"); | |
250 | sbi->s_super->s_flags &= ~MS_RDONLY; | |
251 | } | |
252 | ||
253 | err = nilfs_search_super_root(nilfs, sbi, &ri); | |
254 | if (unlikely(err)) { | |
255 | printk(KERN_ERR "NILFS: error searching super root.\n"); | |
256 | goto failed; | |
257 | } | |
258 | ||
259 | err = nilfs_load_super_root(nilfs, sbi, ri.ri_super_root); | |
260 | if (unlikely(err)) { | |
261 | printk(KERN_ERR "NILFS: error loading super root.\n"); | |
262 | goto failed; | |
263 | } | |
264 | ||
265 | if (!valid_fs) { | |
266 | err = nilfs_recover_logical_segments(nilfs, sbi, &ri); | |
267 | if (unlikely(err)) { | |
268 | nilfs_mdt_destroy(nilfs->ns_cpfile); | |
269 | nilfs_mdt_destroy(nilfs->ns_sufile); | |
270 | nilfs_mdt_destroy(nilfs->ns_dat); | |
271 | goto failed; | |
272 | } | |
273 | if (ri.ri_need_recovery == NILFS_RECOVERY_SR_UPDATED) { | |
274 | down_write(&nilfs->ns_sem); | |
275 | nilfs_update_last_segment(sbi, 0); | |
276 | up_write(&nilfs->ns_sem); | |
277 | } | |
278 | } | |
279 | ||
280 | set_nilfs_loaded(nilfs); | |
281 | ||
282 | failed: | |
283 | nilfs_clear_recovery_info(&ri); | |
284 | sbi->s_super->s_flags = s_flags; | |
285 | return err; | |
286 | } | |
287 | ||
288 | static unsigned long long nilfs_max_size(unsigned int blkbits) | |
289 | { | |
290 | unsigned int max_bits; | |
291 | unsigned long long res = MAX_LFS_FILESIZE; /* page cache limit */ | |
292 | ||
293 | max_bits = blkbits + NILFS_BMAP_KEY_BIT; /* bmap size limit */ | |
294 | if (max_bits < 64) | |
295 | res = min_t(unsigned long long, res, (1ULL << max_bits) - 1); | |
296 | return res; | |
297 | } | |
298 | ||
299 | static int | |
300 | nilfs_store_disk_layout(struct the_nilfs *nilfs, struct super_block *sb, | |
301 | struct nilfs_super_block *sbp) | |
302 | { | |
303 | if (le32_to_cpu(sbp->s_rev_level) != NILFS_CURRENT_REV) { | |
304 | printk(KERN_ERR "NILFS: revision mismatch " | |
305 | "(superblock rev.=%d.%d, current rev.=%d.%d). " | |
306 | "Please check the version of mkfs.nilfs.\n", | |
307 | le32_to_cpu(sbp->s_rev_level), | |
308 | le16_to_cpu(sbp->s_minor_rev_level), | |
309 | NILFS_CURRENT_REV, NILFS_MINOR_REV); | |
310 | return -EINVAL; | |
311 | } | |
312 | nilfs->ns_inode_size = le16_to_cpu(sbp->s_inode_size); | |
313 | nilfs->ns_first_ino = le32_to_cpu(sbp->s_first_ino); | |
314 | ||
315 | nilfs->ns_blocks_per_segment = le32_to_cpu(sbp->s_blocks_per_segment); | |
316 | if (nilfs->ns_blocks_per_segment < NILFS_SEG_MIN_BLOCKS) { | |
317 | printk(KERN_ERR "NILFS: too short segment. \n"); | |
318 | return -EINVAL; | |
319 | } | |
320 | ||
321 | nilfs->ns_first_data_block = le64_to_cpu(sbp->s_first_data_block); | |
322 | nilfs->ns_nsegments = le64_to_cpu(sbp->s_nsegments); | |
323 | nilfs->ns_r_segments_percentage = | |
324 | le32_to_cpu(sbp->s_r_segments_percentage); | |
325 | nilfs->ns_nrsvsegs = | |
326 | max_t(unsigned long, NILFS_MIN_NRSVSEGS, | |
327 | DIV_ROUND_UP(nilfs->ns_nsegments * | |
328 | nilfs->ns_r_segments_percentage, 100)); | |
329 | nilfs->ns_crc_seed = le32_to_cpu(sbp->s_crc_seed); | |
330 | return 0; | |
331 | } | |
332 | ||
333 | /** | |
334 | * init_nilfs - initialize a NILFS instance. | |
335 | * @nilfs: the_nilfs structure | |
336 | * @sbi: nilfs_sb_info | |
337 | * @sb: super block | |
338 | * @data: mount options | |
339 | * | |
340 | * init_nilfs() performs common initialization per block device (e.g. | |
341 | * reading the super block, getting disk layout information, initializing | |
342 | * shared fields in the_nilfs). It takes on some portion of the jobs | |
343 | * typically done by a fill_super() routine. This division arises from | |
344 | * the nature that multiple NILFS instances may be simultaneously | |
345 | * mounted on a device. | |
346 | * For multiple mounts on the same device, only the first mount | |
347 | * invokes these tasks. | |
348 | * | |
349 | * Return Value: On success, 0 is returned. On error, a negative error | |
350 | * code is returned. | |
351 | */ | |
352 | int init_nilfs(struct the_nilfs *nilfs, struct nilfs_sb_info *sbi, char *data) | |
353 | { | |
354 | struct super_block *sb = sbi->s_super; | |
355 | struct buffer_head *sbh; | |
356 | struct nilfs_super_block *sbp; | |
357 | struct backing_dev_info *bdi; | |
358 | int blocksize; | |
359 | int err = 0; | |
360 | ||
361 | down_write(&nilfs->ns_sem); | |
362 | if (nilfs_init(nilfs)) { | |
363 | /* Load values from existing the_nilfs */ | |
364 | sbp = nilfs->ns_sbp; | |
365 | err = nilfs_store_magic_and_option(sb, sbp, data); | |
366 | if (err) | |
367 | goto out; | |
368 | ||
369 | blocksize = BLOCK_SIZE << le32_to_cpu(sbp->s_log_block_size); | |
370 | if (sb->s_blocksize != blocksize && | |
371 | !sb_set_blocksize(sb, blocksize)) { | |
372 | printk(KERN_ERR "NILFS: blocksize %d unfit to device\n", | |
373 | blocksize); | |
374 | err = -EINVAL; | |
375 | } | |
376 | sb->s_maxbytes = nilfs_max_size(sb->s_blocksize_bits); | |
377 | goto out; | |
378 | } | |
379 | ||
380 | sbp = nilfs_load_super_block(sb, &sbh); | |
381 | if (!sbp) { | |
382 | err = -EINVAL; | |
383 | goto out; | |
384 | } | |
385 | err = nilfs_store_magic_and_option(sb, sbp, data); | |
386 | if (err) | |
387 | goto failed_sbh; | |
388 | ||
389 | blocksize = BLOCK_SIZE << le32_to_cpu(sbp->s_log_block_size); | |
390 | if (sb->s_blocksize != blocksize) { | |
391 | sbp = nilfs_reload_super_block(sb, &sbh, blocksize); | |
392 | if (!sbp) { | |
393 | err = -EINVAL; | |
394 | goto out; | |
395 | /* not failed_sbh; sbh is released automatically | |
396 | when reloading fails. */ | |
397 | } | |
398 | } | |
399 | nilfs->ns_blocksize_bits = sb->s_blocksize_bits; | |
400 | ||
401 | err = nilfs_store_disk_layout(nilfs, sb, sbp); | |
402 | if (err) | |
403 | goto failed_sbh; | |
404 | ||
405 | sb->s_maxbytes = nilfs_max_size(sb->s_blocksize_bits); | |
406 | ||
407 | nilfs->ns_mount_state = le16_to_cpu(sbp->s_state); | |
408 | nilfs->ns_sbh = sbh; | |
409 | nilfs->ns_sbp = sbp; | |
410 | ||
411 | bdi = nilfs->ns_bdev->bd_inode_backing_dev_info; | |
412 | if (!bdi) | |
413 | bdi = nilfs->ns_bdev->bd_inode->i_mapping->backing_dev_info; | |
414 | nilfs->ns_bdi = bdi ? : &default_backing_dev_info; | |
415 | ||
416 | /* Finding last segment */ | |
417 | nilfs->ns_last_pseg = le64_to_cpu(sbp->s_last_pseg); | |
418 | nilfs->ns_last_cno = le64_to_cpu(sbp->s_last_cno); | |
419 | nilfs->ns_last_seq = le64_to_cpu(sbp->s_last_seq); | |
420 | ||
421 | nilfs->ns_seg_seq = nilfs->ns_last_seq; | |
422 | nilfs->ns_segnum = | |
423 | nilfs_get_segnum_of_block(nilfs, nilfs->ns_last_pseg); | |
424 | nilfs->ns_cno = nilfs->ns_last_cno + 1; | |
425 | if (nilfs->ns_segnum >= nilfs->ns_nsegments) { | |
426 | printk(KERN_ERR "NILFS invalid last segment number.\n"); | |
427 | err = -EINVAL; | |
428 | goto failed_sbh; | |
429 | } | |
430 | /* Dummy values */ | |
431 | nilfs->ns_free_segments_count = | |
432 | nilfs->ns_nsegments - (nilfs->ns_segnum + 1); | |
433 | ||
434 | /* Initialize gcinode cache */ | |
435 | err = nilfs_init_gccache(nilfs); | |
436 | if (err) | |
437 | goto failed_sbh; | |
438 | ||
439 | set_nilfs_init(nilfs); | |
440 | err = 0; | |
441 | out: | |
442 | up_write(&nilfs->ns_sem); | |
443 | return err; | |
444 | ||
445 | failed_sbh: | |
446 | brelse(sbh); | |
447 | goto out; | |
448 | } | |
449 | ||
450 | int nilfs_count_free_blocks(struct the_nilfs *nilfs, sector_t *nblocks) | |
451 | { | |
452 | struct inode *dat = nilfs_dat_inode(nilfs); | |
453 | unsigned long ncleansegs; | |
454 | int err; | |
455 | ||
456 | down_read(&NILFS_MDT(dat)->mi_sem); /* XXX */ | |
457 | err = nilfs_sufile_get_ncleansegs(nilfs->ns_sufile, &ncleansegs); | |
458 | up_read(&NILFS_MDT(dat)->mi_sem); /* XXX */ | |
459 | if (likely(!err)) | |
460 | *nblocks = (sector_t)ncleansegs * nilfs->ns_blocks_per_segment; | |
461 | return err; | |
462 | } | |
463 | ||
8a9d2191 RK |
464 | int nilfs_near_disk_full(struct the_nilfs *nilfs) |
465 | { | |
466 | struct inode *sufile = nilfs->ns_sufile; | |
467 | unsigned long ncleansegs, nincsegs; | |
468 | int ret; | |
469 | ||
470 | ret = nilfs_sufile_get_ncleansegs(sufile, &ncleansegs); | |
471 | if (likely(!ret)) { | |
472 | nincsegs = atomic_read(&nilfs->ns_ndirtyblks) / | |
473 | nilfs->ns_blocks_per_segment + 1; | |
474 | if (ncleansegs <= nilfs->ns_nrsvsegs + nincsegs) | |
475 | ret++; | |
476 | } | |
477 | return ret; | |
478 | } | |
479 | ||
480 | int nilfs_checkpoint_is_mounted(struct the_nilfs *nilfs, __u64 cno, | |
481 | int snapshot_mount) | |
482 | { | |
483 | struct nilfs_sb_info *sbi; | |
484 | int ret = 0; | |
485 | ||
486 | down_read(&nilfs->ns_sem); | |
487 | if (cno == 0 || cno > nilfs->ns_cno) | |
488 | goto out_unlock; | |
489 | ||
490 | list_for_each_entry(sbi, &nilfs->ns_supers, s_list) { | |
491 | if (sbi->s_snapshot_cno == cno && | |
492 | (!snapshot_mount || nilfs_test_opt(sbi, SNAPSHOT))) { | |
493 | /* exclude read-only mounts */ | |
494 | ret++; | |
495 | break; | |
496 | } | |
497 | } | |
498 | /* for protecting recent checkpoints */ | |
499 | if (cno >= nilfs_last_cno(nilfs)) | |
500 | ret++; | |
501 | ||
502 | out_unlock: | |
503 | up_read(&nilfs->ns_sem); | |
504 | return ret; | |
505 | } |