blob: 59e5fe742f7bd7bcc6d6aa88dbdcf595e4c64019 [file] [log] [blame]
Koji Satoa17564f2009-04-06 19:01:30 -07001/*
2 * dat.c - NILFS disk address translation.
3 *
4 * Copyright (C) 2006-2008 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 Koji Sato <koji@osrg.net>.
21 */
22
23#include <linux/types.h>
24#include <linux/buffer_head.h>
25#include <linux/string.h>
26#include <linux/errno.h>
27#include "nilfs.h"
28#include "mdt.h"
29#include "alloc.h"
30#include "dat.h"
31
32
33#define NILFS_CNO_MIN ((__u64)1)
34#define NILFS_CNO_MAX (~(__u64)0)
35
Ryusuke Konishi8908b2f2009-11-14 16:35:01 +090036struct nilfs_dat_info {
37 struct nilfs_mdt_info mi;
38 struct nilfs_palloc_cache palloc_cache;
Ryusuke Konishic1c1d70922010-08-29 12:44:56 +090039 struct nilfs_shadow_map shadow;
Ryusuke Konishi8908b2f2009-11-14 16:35:01 +090040};
41
42static inline struct nilfs_dat_info *NILFS_DAT_I(struct inode *dat)
43{
44 return (struct nilfs_dat_info *)NILFS_MDT(dat);
45}
46
Koji Satoa17564f2009-04-06 19:01:30 -070047static int nilfs_dat_prepare_entry(struct inode *dat,
48 struct nilfs_palloc_req *req, int create)
49{
50 return nilfs_palloc_get_entry_block(dat, req->pr_entry_nr,
51 create, &req->pr_entry_bh);
52}
53
54static void nilfs_dat_commit_entry(struct inode *dat,
55 struct nilfs_palloc_req *req)
56{
57 nilfs_mdt_mark_buffer_dirty(req->pr_entry_bh);
58 nilfs_mdt_mark_dirty(dat);
59 brelse(req->pr_entry_bh);
60}
61
62static void nilfs_dat_abort_entry(struct inode *dat,
63 struct nilfs_palloc_req *req)
64{
65 brelse(req->pr_entry_bh);
66}
67
68int nilfs_dat_prepare_alloc(struct inode *dat, struct nilfs_palloc_req *req)
69{
70 int ret;
71
72 ret = nilfs_palloc_prepare_alloc_entry(dat, req);
73 if (ret < 0)
74 return ret;
75
76 ret = nilfs_dat_prepare_entry(dat, req, 1);
77 if (ret < 0)
78 nilfs_palloc_abort_alloc_entry(dat, req);
79
80 return ret;
81}
82
83void nilfs_dat_commit_alloc(struct inode *dat, struct nilfs_palloc_req *req)
84{
85 struct nilfs_dat_entry *entry;
86 void *kaddr;
87
88 kaddr = kmap_atomic(req->pr_entry_bh->b_page, KM_USER0);
89 entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
90 req->pr_entry_bh, kaddr);
91 entry->de_start = cpu_to_le64(NILFS_CNO_MIN);
92 entry->de_end = cpu_to_le64(NILFS_CNO_MAX);
93 entry->de_blocknr = cpu_to_le64(0);
94 kunmap_atomic(kaddr, KM_USER0);
95
96 nilfs_palloc_commit_alloc_entry(dat, req);
97 nilfs_dat_commit_entry(dat, req);
98}
99
100void nilfs_dat_abort_alloc(struct inode *dat, struct nilfs_palloc_req *req)
101{
102 nilfs_dat_abort_entry(dat, req);
103 nilfs_palloc_abort_alloc_entry(dat, req);
104}
105
Jiro SEKIBAabc0b502010-10-08 22:37:27 +0900106static void nilfs_dat_commit_free(struct inode *dat,
107 struct nilfs_palloc_req *req)
Koji Satoa17564f2009-04-06 19:01:30 -0700108{
109 struct nilfs_dat_entry *entry;
110 void *kaddr;
111
112 kaddr = kmap_atomic(req->pr_entry_bh->b_page, KM_USER0);
113 entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
114 req->pr_entry_bh, kaddr);
115 entry->de_start = cpu_to_le64(NILFS_CNO_MIN);
116 entry->de_end = cpu_to_le64(NILFS_CNO_MIN);
117 entry->de_blocknr = cpu_to_le64(0);
118 kunmap_atomic(kaddr, KM_USER0);
119
120 nilfs_dat_commit_entry(dat, req);
121 nilfs_palloc_commit_free_entry(dat, req);
122}
123
Koji Satoa17564f2009-04-06 19:01:30 -0700124int nilfs_dat_prepare_start(struct inode *dat, struct nilfs_palloc_req *req)
125{
126 int ret;
127
128 ret = nilfs_dat_prepare_entry(dat, req, 0);
Ryusuke Konishi1f5abe72009-04-06 19:01:55 -0700129 WARN_ON(ret == -ENOENT);
Koji Satoa17564f2009-04-06 19:01:30 -0700130 return ret;
131}
132
133void nilfs_dat_commit_start(struct inode *dat, struct nilfs_palloc_req *req,
134 sector_t blocknr)
135{
136 struct nilfs_dat_entry *entry;
137 void *kaddr;
138
139 kaddr = kmap_atomic(req->pr_entry_bh->b_page, KM_USER0);
140 entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
141 req->pr_entry_bh, kaddr);
142 entry->de_start = cpu_to_le64(nilfs_mdt_cno(dat));
Koji Satoa17564f2009-04-06 19:01:30 -0700143 entry->de_blocknr = cpu_to_le64(blocknr);
144 kunmap_atomic(kaddr, KM_USER0);
145
146 nilfs_dat_commit_entry(dat, req);
147}
148
Koji Satoa17564f2009-04-06 19:01:30 -0700149int nilfs_dat_prepare_end(struct inode *dat, struct nilfs_palloc_req *req)
150{
151 struct nilfs_dat_entry *entry;
152 __u64 start;
153 sector_t blocknr;
154 void *kaddr;
155 int ret;
156
157 ret = nilfs_dat_prepare_entry(dat, req, 0);
158 if (ret < 0) {
Ryusuke Konishi1f5abe72009-04-06 19:01:55 -0700159 WARN_ON(ret == -ENOENT);
Koji Satoa17564f2009-04-06 19:01:30 -0700160 return ret;
161 }
162
163 kaddr = kmap_atomic(req->pr_entry_bh->b_page, KM_USER0);
164 entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
165 req->pr_entry_bh, kaddr);
166 start = le64_to_cpu(entry->de_start);
167 blocknr = le64_to_cpu(entry->de_blocknr);
168 kunmap_atomic(kaddr, KM_USER0);
169
170 if (blocknr == 0) {
171 ret = nilfs_palloc_prepare_free_entry(dat, req);
172 if (ret < 0) {
173 nilfs_dat_abort_entry(dat, req);
174 return ret;
175 }
176 }
177
178 return 0;
179}
180
181void nilfs_dat_commit_end(struct inode *dat, struct nilfs_palloc_req *req,
182 int dead)
183{
184 struct nilfs_dat_entry *entry;
185 __u64 start, end;
186 sector_t blocknr;
187 void *kaddr;
188
189 kaddr = kmap_atomic(req->pr_entry_bh->b_page, KM_USER0);
190 entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
191 req->pr_entry_bh, kaddr);
192 end = start = le64_to_cpu(entry->de_start);
193 if (!dead) {
194 end = nilfs_mdt_cno(dat);
Ryusuke Konishi1f5abe72009-04-06 19:01:55 -0700195 WARN_ON(start > end);
Koji Satoa17564f2009-04-06 19:01:30 -0700196 }
197 entry->de_end = cpu_to_le64(end);
198 blocknr = le64_to_cpu(entry->de_blocknr);
199 kunmap_atomic(kaddr, KM_USER0);
200
201 if (blocknr == 0)
202 nilfs_dat_commit_free(dat, req);
203 else
204 nilfs_dat_commit_entry(dat, req);
205}
206
207void nilfs_dat_abort_end(struct inode *dat, struct nilfs_palloc_req *req)
208{
209 struct nilfs_dat_entry *entry;
210 __u64 start;
211 sector_t blocknr;
212 void *kaddr;
213
214 kaddr = kmap_atomic(req->pr_entry_bh->b_page, KM_USER0);
215 entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
216 req->pr_entry_bh, kaddr);
217 start = le64_to_cpu(entry->de_start);
218 blocknr = le64_to_cpu(entry->de_blocknr);
219 kunmap_atomic(kaddr, KM_USER0);
220
221 if (start == nilfs_mdt_cno(dat) && blocknr == 0)
222 nilfs_palloc_abort_free_entry(dat, req);
223 nilfs_dat_abort_entry(dat, req);
224}
225
Ryusuke Konishibd8169e2009-08-15 17:22:13 +0900226int nilfs_dat_prepare_update(struct inode *dat,
227 struct nilfs_palloc_req *oldreq,
228 struct nilfs_palloc_req *newreq)
229{
230 int ret;
231
232 ret = nilfs_dat_prepare_end(dat, oldreq);
233 if (!ret) {
234 ret = nilfs_dat_prepare_alloc(dat, newreq);
235 if (ret < 0)
236 nilfs_dat_abort_end(dat, oldreq);
237 }
238 return ret;
239}
240
241void nilfs_dat_commit_update(struct inode *dat,
242 struct nilfs_palloc_req *oldreq,
243 struct nilfs_palloc_req *newreq, int dead)
244{
245 nilfs_dat_commit_end(dat, oldreq, dead);
246 nilfs_dat_commit_alloc(dat, newreq);
247}
248
249void nilfs_dat_abort_update(struct inode *dat,
250 struct nilfs_palloc_req *oldreq,
251 struct nilfs_palloc_req *newreq)
252{
253 nilfs_dat_abort_end(dat, oldreq);
254 nilfs_dat_abort_alloc(dat, newreq);
255}
256
Koji Satoa17564f2009-04-06 19:01:30 -0700257/**
258 * nilfs_dat_mark_dirty -
259 * @dat: DAT file inode
260 * @vblocknr: virtual block number
261 *
262 * Description:
263 *
264 * Return Value: On success, 0 is returned. On error, one of the following
265 * negative error codes is returned.
266 *
267 * %-EIO - I/O error.
268 *
269 * %-ENOMEM - Insufficient amount of memory available.
270 */
271int nilfs_dat_mark_dirty(struct inode *dat, __u64 vblocknr)
272{
273 struct nilfs_palloc_req req;
274 int ret;
275
276 req.pr_entry_nr = vblocknr;
277 ret = nilfs_dat_prepare_entry(dat, &req, 0);
278 if (ret == 0)
279 nilfs_dat_commit_entry(dat, &req);
280 return ret;
281}
282
283/**
284 * nilfs_dat_freev - free virtual block numbers
285 * @dat: DAT file inode
286 * @vblocknrs: array of virtual block numbers
287 * @nitems: number of virtual block numbers
288 *
289 * Description: nilfs_dat_freev() frees the virtual block numbers specified by
290 * @vblocknrs and @nitems.
291 *
292 * Return Value: On success, 0 is returned. On error, one of the following
Ryusuke Konishi7a650042010-03-14 03:32:40 +0900293 * negative error codes is returned.
Koji Satoa17564f2009-04-06 19:01:30 -0700294 *
295 * %-EIO - I/O error.
296 *
297 * %-ENOMEM - Insufficient amount of memory available.
298 *
299 * %-ENOENT - The virtual block number have not been allocated.
300 */
301int nilfs_dat_freev(struct inode *dat, __u64 *vblocknrs, size_t nitems)
302{
303 return nilfs_palloc_freev(dat, vblocknrs, nitems);
304}
305
306/**
307 * nilfs_dat_move - change a block number
308 * @dat: DAT file inode
309 * @vblocknr: virtual block number
310 * @blocknr: block number
311 *
312 * Description: nilfs_dat_move() changes the block number associated with
313 * @vblocknr to @blocknr.
314 *
315 * Return Value: On success, 0 is returned. On error, one of the following
316 * negative error codes is returned.
317 *
318 * %-EIO - I/O error.
319 *
320 * %-ENOMEM - Insufficient amount of memory available.
321 */
322int nilfs_dat_move(struct inode *dat, __u64 vblocknr, sector_t blocknr)
323{
324 struct buffer_head *entry_bh;
325 struct nilfs_dat_entry *entry;
326 void *kaddr;
327 int ret;
328
329 ret = nilfs_palloc_get_entry_block(dat, vblocknr, 0, &entry_bh);
330 if (ret < 0)
331 return ret;
Ryusuke Konishic1c1d70922010-08-29 12:44:56 +0900332
333 /*
334 * The given disk block number (blocknr) is not yet written to
335 * the device at this point.
336 *
337 * To prevent nilfs_dat_translate() from returning the
Ryusuke Konishif6c26ec2010-11-24 02:18:59 +0900338 * uncommitted block number, this makes a copy of the entry
Ryusuke Konishic1c1d70922010-08-29 12:44:56 +0900339 * buffer and redirects nilfs_dat_translate() to the copy.
340 */
341 if (!buffer_nilfs_redirected(entry_bh)) {
342 ret = nilfs_mdt_freeze_buffer(dat, entry_bh);
343 if (ret) {
344 brelse(entry_bh);
345 return ret;
346 }
347 }
348
Koji Satoa17564f2009-04-06 19:01:30 -0700349 kaddr = kmap_atomic(entry_bh->b_page, KM_USER0);
350 entry = nilfs_palloc_block_get_entry(dat, vblocknr, entry_bh, kaddr);
Ryusuke Konishi1f5abe72009-04-06 19:01:55 -0700351 if (unlikely(entry->de_blocknr == cpu_to_le64(0))) {
Koji Satoa17564f2009-04-06 19:01:30 -0700352 printk(KERN_CRIT "%s: vbn = %llu, [%llu, %llu)\n", __func__,
353 (unsigned long long)vblocknr,
354 (unsigned long long)le64_to_cpu(entry->de_start),
355 (unsigned long long)le64_to_cpu(entry->de_end));
Ryusuke Konishi1f5abe72009-04-06 19:01:55 -0700356 kunmap_atomic(kaddr, KM_USER0);
357 brelse(entry_bh);
358 return -EINVAL;
Koji Satoa17564f2009-04-06 19:01:30 -0700359 }
Ryusuke Konishi1f5abe72009-04-06 19:01:55 -0700360 WARN_ON(blocknr == 0);
Koji Satoa17564f2009-04-06 19:01:30 -0700361 entry->de_blocknr = cpu_to_le64(blocknr);
362 kunmap_atomic(kaddr, KM_USER0);
363
364 nilfs_mdt_mark_buffer_dirty(entry_bh);
365 nilfs_mdt_mark_dirty(dat);
366
367 brelse(entry_bh);
368
369 return 0;
370}
371
372/**
373 * nilfs_dat_translate - translate a virtual block number to a block number
374 * @dat: DAT file inode
375 * @vblocknr: virtual block number
376 * @blocknrp: pointer to a block number
377 *
378 * Description: nilfs_dat_translate() maps the virtual block number @vblocknr
379 * to the corresponding block number.
380 *
381 * Return Value: On success, 0 is returned and the block number associated
382 * with @vblocknr is stored in the place pointed by @blocknrp. On error, one
383 * of the following negative error codes is returned.
384 *
385 * %-EIO - I/O error.
386 *
387 * %-ENOMEM - Insufficient amount of memory available.
388 *
389 * %-ENOENT - A block number associated with @vblocknr does not exist.
390 */
391int nilfs_dat_translate(struct inode *dat, __u64 vblocknr, sector_t *blocknrp)
392{
Ryusuke Konishic1c1d70922010-08-29 12:44:56 +0900393 struct buffer_head *entry_bh, *bh;
Koji Satoa17564f2009-04-06 19:01:30 -0700394 struct nilfs_dat_entry *entry;
395 sector_t blocknr;
396 void *kaddr;
397 int ret;
398
399 ret = nilfs_palloc_get_entry_block(dat, vblocknr, 0, &entry_bh);
400 if (ret < 0)
401 return ret;
402
Ryusuke Konishic1c1d70922010-08-29 12:44:56 +0900403 if (!nilfs_doing_gc() && buffer_nilfs_redirected(entry_bh)) {
404 bh = nilfs_mdt_get_frozen_buffer(dat, entry_bh);
405 if (bh) {
406 WARN_ON(!buffer_uptodate(bh));
407 brelse(entry_bh);
408 entry_bh = bh;
409 }
410 }
411
Koji Satoa17564f2009-04-06 19:01:30 -0700412 kaddr = kmap_atomic(entry_bh->b_page, KM_USER0);
413 entry = nilfs_palloc_block_get_entry(dat, vblocknr, entry_bh, kaddr);
414 blocknr = le64_to_cpu(entry->de_blocknr);
415 if (blocknr == 0) {
416 ret = -ENOENT;
417 goto out;
418 }
Jiro SEKIBA086d1762010-02-05 23:15:26 +0900419 *blocknrp = blocknr;
Koji Satoa17564f2009-04-06 19:01:30 -0700420
421 out:
422 kunmap_atomic(kaddr, KM_USER0);
423 brelse(entry_bh);
424 return ret;
425}
426
Ryusuke Konishi003ff182009-05-12 03:58:47 +0900427ssize_t nilfs_dat_get_vinfo(struct inode *dat, void *buf, unsigned visz,
Koji Satoa17564f2009-04-06 19:01:30 -0700428 size_t nvi)
429{
430 struct buffer_head *entry_bh;
431 struct nilfs_dat_entry *entry;
Ryusuke Konishi003ff182009-05-12 03:58:47 +0900432 struct nilfs_vinfo *vinfo = buf;
Koji Satoa17564f2009-04-06 19:01:30 -0700433 __u64 first, last;
434 void *kaddr;
435 unsigned long entries_per_block = NILFS_MDT(dat)->mi_entries_per_block;
436 int i, j, n, ret;
437
438 for (i = 0; i < nvi; i += n) {
Ryusuke Konishi003ff182009-05-12 03:58:47 +0900439 ret = nilfs_palloc_get_entry_block(dat, vinfo->vi_vblocknr,
Koji Satoa17564f2009-04-06 19:01:30 -0700440 0, &entry_bh);
441 if (ret < 0)
442 return ret;
443 kaddr = kmap_atomic(entry_bh->b_page, KM_USER0);
444 /* last virtual block number in this block */
Ryusuke Konishi003ff182009-05-12 03:58:47 +0900445 first = vinfo->vi_vblocknr;
Koji Satoa17564f2009-04-06 19:01:30 -0700446 do_div(first, entries_per_block);
447 first *= entries_per_block;
448 last = first + entries_per_block - 1;
449 for (j = i, n = 0;
Ryusuke Konishi003ff182009-05-12 03:58:47 +0900450 j < nvi && vinfo->vi_vblocknr >= first &&
451 vinfo->vi_vblocknr <= last;
452 j++, n++, vinfo = (void *)vinfo + visz) {
Koji Satoa17564f2009-04-06 19:01:30 -0700453 entry = nilfs_palloc_block_get_entry(
Ryusuke Konishi003ff182009-05-12 03:58:47 +0900454 dat, vinfo->vi_vblocknr, entry_bh, kaddr);
455 vinfo->vi_start = le64_to_cpu(entry->de_start);
456 vinfo->vi_end = le64_to_cpu(entry->de_end);
457 vinfo->vi_blocknr = le64_to_cpu(entry->de_blocknr);
Koji Satoa17564f2009-04-06 19:01:30 -0700458 }
459 kunmap_atomic(kaddr, KM_USER0);
460 brelse(entry_bh);
461 }
462
463 return nvi;
464}
Ryusuke Konishi79739562009-11-12 23:56:43 +0900465
466/**
Ryusuke Konishif1e89c82010-09-05 12:20:59 +0900467 * nilfs_dat_read - read or get dat inode
468 * @sb: super block instance
Ryusuke Konishi79739562009-11-12 23:56:43 +0900469 * @entry_size: size of a dat entry
Ryusuke Konishif1e89c82010-09-05 12:20:59 +0900470 * @raw_inode: on-disk dat inode
471 * @inodep: buffer to store the inode
Ryusuke Konishi79739562009-11-12 23:56:43 +0900472 */
Ryusuke Konishif1e89c82010-09-05 12:20:59 +0900473int nilfs_dat_read(struct super_block *sb, size_t entry_size,
474 struct nilfs_inode *raw_inode, struct inode **inodep)
Ryusuke Konishi79739562009-11-12 23:56:43 +0900475{
476 static struct lock_class_key dat_lock_key;
477 struct inode *dat;
Ryusuke Konishi8908b2f2009-11-14 16:35:01 +0900478 struct nilfs_dat_info *di;
Ryusuke Konishi79739562009-11-12 23:56:43 +0900479 int err;
480
Ryusuke Konishif1e89c82010-09-05 12:20:59 +0900481 dat = nilfs_iget_locked(sb, NULL, NILFS_DAT_INO);
482 if (unlikely(!dat))
483 return -ENOMEM;
484 if (!(dat->i_state & I_NEW))
485 goto out;
Ryusuke Konishi8908b2f2009-11-14 16:35:01 +0900486
Ryusuke Konishif1e89c82010-09-05 12:20:59 +0900487 err = nilfs_mdt_init(dat, NILFS_MDT_GFP, sizeof(*di));
488 if (err)
489 goto failed;
490
491 err = nilfs_palloc_init_blockgroup(dat, entry_size);
492 if (err)
493 goto failed;
494
495 di = NILFS_DAT_I(dat);
496 lockdep_set_class(&di->mi.mi_sem, &dat_lock_key);
497 nilfs_palloc_setup_cache(dat, &di->palloc_cache);
498 nilfs_mdt_setup_shadow_map(dat, &di->shadow);
499
500 err = nilfs_read_inode_common(dat, raw_inode);
501 if (err)
502 goto failed;
503
504 unlock_new_inode(dat);
505 out:
506 *inodep = dat;
507 return 0;
508 failed:
509 iget_failed(dat);
510 return err;
Ryusuke Konishi79739562009-11-12 23:56:43 +0900511}