Commit | Line | Data |
---|---|---|
64b5a32e RK |
1 | /* |
2 | * segbuf.c - NILFS segment buffer | |
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/writeback.h> | |
26 | #include <linux/crc32.h> | |
27 | #include "page.h" | |
28 | #include "segbuf.h" | |
64b5a32e RK |
29 | |
30 | ||
31 | static struct kmem_cache *nilfs_segbuf_cachep; | |
32 | ||
33 | static void nilfs_segbuf_init_once(void *obj) | |
34 | { | |
35 | memset(obj, 0, sizeof(struct nilfs_segment_buffer)); | |
36 | } | |
37 | ||
38 | int __init nilfs_init_segbuf_cache(void) | |
39 | { | |
40 | nilfs_segbuf_cachep = | |
41 | kmem_cache_create("nilfs2_segbuf_cache", | |
42 | sizeof(struct nilfs_segment_buffer), | |
43 | 0, SLAB_RECLAIM_ACCOUNT, | |
44 | nilfs_segbuf_init_once); | |
45 | ||
46 | return (nilfs_segbuf_cachep == NULL) ? -ENOMEM : 0; | |
47 | } | |
48 | ||
49 | void nilfs_destroy_segbuf_cache(void) | |
50 | { | |
51 | kmem_cache_destroy(nilfs_segbuf_cachep); | |
52 | } | |
53 | ||
54 | struct nilfs_segment_buffer *nilfs_segbuf_new(struct super_block *sb) | |
55 | { | |
56 | struct nilfs_segment_buffer *segbuf; | |
57 | ||
58 | segbuf = kmem_cache_alloc(nilfs_segbuf_cachep, GFP_NOFS); | |
59 | if (unlikely(!segbuf)) | |
60 | return NULL; | |
61 | ||
62 | segbuf->sb_super = sb; | |
63 | INIT_LIST_HEAD(&segbuf->sb_list); | |
64 | INIT_LIST_HEAD(&segbuf->sb_segsum_buffers); | |
65 | INIT_LIST_HEAD(&segbuf->sb_payload_buffers); | |
64b5a32e RK |
66 | return segbuf; |
67 | } | |
68 | ||
69 | void nilfs_segbuf_free(struct nilfs_segment_buffer *segbuf) | |
70 | { | |
64b5a32e RK |
71 | kmem_cache_free(nilfs_segbuf_cachep, segbuf); |
72 | } | |
73 | ||
cece5520 | 74 | void nilfs_segbuf_map(struct nilfs_segment_buffer *segbuf, __u64 segnum, |
64b5a32e RK |
75 | unsigned long offset, struct the_nilfs *nilfs) |
76 | { | |
64b5a32e RK |
77 | segbuf->sb_segnum = segnum; |
78 | nilfs_get_segment_range(nilfs, segnum, &segbuf->sb_fseg_start, | |
79 | &segbuf->sb_fseg_end); | |
80 | ||
81 | segbuf->sb_pseg_start = segbuf->sb_fseg_start + offset; | |
82 | segbuf->sb_rest_blocks = | |
83 | segbuf->sb_fseg_end - segbuf->sb_pseg_start + 1; | |
64b5a32e RK |
84 | } |
85 | ||
86 | void nilfs_segbuf_set_next_segnum(struct nilfs_segment_buffer *segbuf, | |
87 | __u64 nextnum, struct the_nilfs *nilfs) | |
88 | { | |
89 | segbuf->sb_nextnum = nextnum; | |
90 | segbuf->sb_sum.next = nilfs_get_segment_start_blocknr(nilfs, nextnum); | |
91 | } | |
92 | ||
93 | int nilfs_segbuf_extend_segsum(struct nilfs_segment_buffer *segbuf) | |
94 | { | |
95 | struct buffer_head *bh; | |
96 | ||
97 | bh = sb_getblk(segbuf->sb_super, | |
98 | segbuf->sb_pseg_start + segbuf->sb_sum.nsumblk); | |
99 | if (unlikely(!bh)) | |
100 | return -ENOMEM; | |
101 | ||
102 | nilfs_segbuf_add_segsum_buffer(segbuf, bh); | |
103 | return 0; | |
104 | } | |
105 | ||
106 | int nilfs_segbuf_extend_payload(struct nilfs_segment_buffer *segbuf, | |
107 | struct buffer_head **bhp) | |
108 | { | |
109 | struct buffer_head *bh; | |
110 | ||
111 | bh = sb_getblk(segbuf->sb_super, | |
112 | segbuf->sb_pseg_start + segbuf->sb_sum.nblocks); | |
113 | if (unlikely(!bh)) | |
114 | return -ENOMEM; | |
115 | ||
116 | nilfs_segbuf_add_payload_buffer(segbuf, bh); | |
117 | *bhp = bh; | |
118 | return 0; | |
119 | } | |
120 | ||
121 | int nilfs_segbuf_reset(struct nilfs_segment_buffer *segbuf, unsigned flags, | |
122 | time_t ctime) | |
123 | { | |
124 | int err; | |
125 | ||
126 | segbuf->sb_sum.nblocks = segbuf->sb_sum.nsumblk = 0; | |
127 | err = nilfs_segbuf_extend_segsum(segbuf); | |
128 | if (unlikely(err)) | |
129 | return err; | |
130 | ||
131 | segbuf->sb_sum.flags = flags; | |
132 | segbuf->sb_sum.sumbytes = sizeof(struct nilfs_segment_summary); | |
133 | segbuf->sb_sum.nfinfo = segbuf->sb_sum.nfileblk = 0; | |
134 | segbuf->sb_sum.ctime = ctime; | |
135 | ||
136 | segbuf->sb_io_error = 0; | |
137 | return 0; | |
138 | } | |
139 | ||
140 | /* | |
141 | * Setup segument summary | |
142 | */ | |
143 | void nilfs_segbuf_fill_in_segsum(struct nilfs_segment_buffer *segbuf) | |
144 | { | |
145 | struct nilfs_segment_summary *raw_sum; | |
146 | struct buffer_head *bh_sum; | |
147 | ||
148 | bh_sum = list_entry(segbuf->sb_segsum_buffers.next, | |
149 | struct buffer_head, b_assoc_buffers); | |
150 | raw_sum = (struct nilfs_segment_summary *)bh_sum->b_data; | |
151 | ||
152 | raw_sum->ss_magic = cpu_to_le32(NILFS_SEGSUM_MAGIC); | |
153 | raw_sum->ss_bytes = cpu_to_le16(sizeof(*raw_sum)); | |
154 | raw_sum->ss_flags = cpu_to_le16(segbuf->sb_sum.flags); | |
155 | raw_sum->ss_seq = cpu_to_le64(segbuf->sb_sum.seg_seq); | |
156 | raw_sum->ss_create = cpu_to_le64(segbuf->sb_sum.ctime); | |
157 | raw_sum->ss_next = cpu_to_le64(segbuf->sb_sum.next); | |
158 | raw_sum->ss_nblocks = cpu_to_le32(segbuf->sb_sum.nblocks); | |
159 | raw_sum->ss_nfinfo = cpu_to_le32(segbuf->sb_sum.nfinfo); | |
160 | raw_sum->ss_sumbytes = cpu_to_le32(segbuf->sb_sum.sumbytes); | |
161 | raw_sum->ss_pad = 0; | |
162 | } | |
163 | ||
164 | /* | |
165 | * CRC calculation routines | |
166 | */ | |
167 | void nilfs_segbuf_fill_in_segsum_crc(struct nilfs_segment_buffer *segbuf, | |
168 | u32 seed) | |
169 | { | |
170 | struct buffer_head *bh; | |
171 | struct nilfs_segment_summary *raw_sum; | |
172 | unsigned long size, bytes = segbuf->sb_sum.sumbytes; | |
173 | u32 crc; | |
174 | ||
175 | bh = list_entry(segbuf->sb_segsum_buffers.next, struct buffer_head, | |
176 | b_assoc_buffers); | |
177 | ||
178 | raw_sum = (struct nilfs_segment_summary *)bh->b_data; | |
179 | size = min_t(unsigned long, bytes, bh->b_size); | |
180 | crc = crc32_le(seed, | |
181 | (unsigned char *)raw_sum + | |
182 | sizeof(raw_sum->ss_datasum) + sizeof(raw_sum->ss_sumsum), | |
183 | size - (sizeof(raw_sum->ss_datasum) + | |
184 | sizeof(raw_sum->ss_sumsum))); | |
185 | ||
186 | list_for_each_entry_continue(bh, &segbuf->sb_segsum_buffers, | |
187 | b_assoc_buffers) { | |
188 | bytes -= size; | |
189 | size = min_t(unsigned long, bytes, bh->b_size); | |
190 | crc = crc32_le(crc, bh->b_data, size); | |
191 | } | |
192 | raw_sum->ss_sumsum = cpu_to_le32(crc); | |
193 | } | |
194 | ||
195 | void nilfs_segbuf_fill_in_data_crc(struct nilfs_segment_buffer *segbuf, | |
196 | u32 seed) | |
197 | { | |
198 | struct buffer_head *bh; | |
199 | struct nilfs_segment_summary *raw_sum; | |
200 | void *kaddr; | |
201 | u32 crc; | |
202 | ||
203 | bh = list_entry(segbuf->sb_segsum_buffers.next, struct buffer_head, | |
204 | b_assoc_buffers); | |
205 | raw_sum = (struct nilfs_segment_summary *)bh->b_data; | |
206 | crc = crc32_le(seed, | |
207 | (unsigned char *)raw_sum + sizeof(raw_sum->ss_datasum), | |
208 | bh->b_size - sizeof(raw_sum->ss_datasum)); | |
209 | ||
210 | list_for_each_entry_continue(bh, &segbuf->sb_segsum_buffers, | |
211 | b_assoc_buffers) { | |
212 | crc = crc32_le(crc, bh->b_data, bh->b_size); | |
213 | } | |
214 | list_for_each_entry(bh, &segbuf->sb_payload_buffers, b_assoc_buffers) { | |
215 | kaddr = kmap_atomic(bh->b_page, KM_USER0); | |
216 | crc = crc32_le(crc, kaddr + bh_offset(bh), bh->b_size); | |
217 | kunmap_atomic(kaddr, KM_USER0); | |
218 | } | |
219 | raw_sum->ss_datasum = cpu_to_le32(crc); | |
220 | } | |
221 | ||
222 | void nilfs_release_buffers(struct list_head *list) | |
223 | { | |
224 | struct buffer_head *bh, *n; | |
225 | ||
226 | list_for_each_entry_safe(bh, n, list, b_assoc_buffers) { | |
227 | list_del_init(&bh->b_assoc_buffers); | |
228 | if (buffer_nilfs_allocated(bh)) { | |
229 | struct page *clone_page = bh->b_page; | |
230 | ||
231 | /* remove clone page */ | |
232 | brelse(bh); | |
233 | page_cache_release(clone_page); /* for each bh */ | |
234 | if (page_count(clone_page) <= 2) { | |
235 | lock_page(clone_page); | |
236 | nilfs_free_private_page(clone_page); | |
237 | } | |
238 | continue; | |
239 | } | |
240 | brelse(bh); | |
241 | } | |
242 | } | |
243 | ||
244 | /* | |
245 | * BIO operations | |
246 | */ | |
247 | static void nilfs_end_bio_write(struct bio *bio, int err) | |
248 | { | |
249 | const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); | |
250 | struct nilfs_write_info *wi = bio->bi_private; | |
251 | ||
252 | if (err == -EOPNOTSUPP) { | |
253 | set_bit(BIO_EOPNOTSUPP, &bio->bi_flags); | |
254 | bio_put(bio); | |
255 | /* to be detected by submit_seg_bio() */ | |
256 | } | |
257 | ||
258 | if (!uptodate) | |
259 | atomic_inc(&wi->err); | |
260 | ||
261 | bio_put(bio); | |
262 | complete(&wi->bio_event); | |
263 | } | |
264 | ||
265 | static int nilfs_submit_seg_bio(struct nilfs_write_info *wi, int mode) | |
266 | { | |
267 | struct bio *bio = wi->bio; | |
268 | int err; | |
269 | ||
270 | if (wi->nbio > 0 && bdi_write_congested(wi->bdi)) { | |
271 | wait_for_completion(&wi->bio_event); | |
272 | wi->nbio--; | |
273 | if (unlikely(atomic_read(&wi->err))) { | |
274 | bio_put(bio); | |
275 | err = -EIO; | |
276 | goto failed; | |
277 | } | |
278 | } | |
279 | ||
280 | bio->bi_end_io = nilfs_end_bio_write; | |
281 | bio->bi_private = wi; | |
282 | bio_get(bio); | |
283 | submit_bio(mode, bio); | |
284 | if (bio_flagged(bio, BIO_EOPNOTSUPP)) { | |
285 | bio_put(bio); | |
286 | err = -EOPNOTSUPP; | |
287 | goto failed; | |
288 | } | |
289 | wi->nbio++; | |
290 | bio_put(bio); | |
291 | ||
292 | wi->bio = NULL; | |
293 | wi->rest_blocks -= wi->end - wi->start; | |
294 | wi->nr_vecs = min(wi->max_pages, wi->rest_blocks); | |
295 | wi->start = wi->end; | |
296 | return 0; | |
297 | ||
298 | failed: | |
299 | wi->bio = NULL; | |
300 | return err; | |
301 | } | |
302 | ||
303 | /** | |
304 | * nilfs_alloc_seg_bio - allocate a bio for writing segment. | |
305 | * @sb: super block | |
306 | * @start: beginning disk block number of this BIO. | |
307 | * @nr_vecs: request size of page vector. | |
308 | * | |
309 | * alloc_seg_bio() allocates a new BIO structure and initialize it. | |
310 | * | |
311 | * Return Value: On success, pointer to the struct bio is returned. | |
312 | * On error, NULL is returned. | |
313 | */ | |
314 | static struct bio *nilfs_alloc_seg_bio(struct super_block *sb, sector_t start, | |
315 | int nr_vecs) | |
316 | { | |
317 | struct bio *bio; | |
318 | ||
c1b353f0 | 319 | bio = bio_alloc(GFP_NOIO, nr_vecs); |
64b5a32e RK |
320 | if (bio == NULL) { |
321 | while (!bio && (nr_vecs >>= 1)) | |
c1b353f0 | 322 | bio = bio_alloc(GFP_NOIO, nr_vecs); |
64b5a32e RK |
323 | } |
324 | if (likely(bio)) { | |
325 | bio->bi_bdev = sb->s_bdev; | |
326 | bio->bi_sector = (sector_t)start << (sb->s_blocksize_bits - 9); | |
327 | } | |
328 | return bio; | |
329 | } | |
330 | ||
331 | void nilfs_segbuf_prepare_write(struct nilfs_segment_buffer *segbuf, | |
332 | struct nilfs_write_info *wi) | |
333 | { | |
334 | wi->bio = NULL; | |
335 | wi->rest_blocks = segbuf->sb_sum.nblocks; | |
336 | wi->max_pages = bio_get_nr_vecs(wi->sb->s_bdev); | |
337 | wi->nr_vecs = min(wi->max_pages, wi->rest_blocks); | |
338 | wi->start = wi->end = 0; | |
339 | wi->nbio = 0; | |
340 | wi->blocknr = segbuf->sb_pseg_start; | |
341 | ||
342 | atomic_set(&wi->err, 0); | |
343 | init_completion(&wi->bio_event); | |
344 | } | |
345 | ||
346 | static int nilfs_submit_bh(struct nilfs_write_info *wi, struct buffer_head *bh, | |
347 | int mode) | |
348 | { | |
349 | int len, err; | |
350 | ||
351 | BUG_ON(wi->nr_vecs <= 0); | |
352 | repeat: | |
353 | if (!wi->bio) { | |
354 | wi->bio = nilfs_alloc_seg_bio(wi->sb, wi->blocknr + wi->end, | |
355 | wi->nr_vecs); | |
356 | if (unlikely(!wi->bio)) | |
357 | return -ENOMEM; | |
358 | } | |
359 | ||
360 | len = bio_add_page(wi->bio, bh->b_page, bh->b_size, bh_offset(bh)); | |
361 | if (len == bh->b_size) { | |
362 | wi->end++; | |
363 | return 0; | |
364 | } | |
365 | /* bio is FULL */ | |
366 | err = nilfs_submit_seg_bio(wi, mode); | |
367 | /* never submit current bh */ | |
368 | if (likely(!err)) | |
369 | goto repeat; | |
370 | return err; | |
371 | } | |
372 | ||
373 | int nilfs_segbuf_write(struct nilfs_segment_buffer *segbuf, | |
374 | struct nilfs_write_info *wi) | |
375 | { | |
376 | struct buffer_head *bh; | |
5f1586d0 | 377 | int res = 0, rw = WRITE; |
64b5a32e RK |
378 | |
379 | list_for_each_entry(bh, &segbuf->sb_segsum_buffers, b_assoc_buffers) { | |
380 | res = nilfs_submit_bh(wi, bh, rw); | |
381 | if (unlikely(res)) | |
382 | goto failed_bio; | |
383 | } | |
384 | ||
385 | list_for_each_entry(bh, &segbuf->sb_payload_buffers, b_assoc_buffers) { | |
386 | res = nilfs_submit_bh(wi, bh, rw); | |
387 | if (unlikely(res)) | |
388 | goto failed_bio; | |
389 | } | |
390 | ||
391 | if (wi->bio) { | |
392 | /* | |
393 | * Last BIO is always sent through the following | |
394 | * submission. | |
395 | */ | |
30bda0b8 | 396 | rw |= (1 << BIO_RW_SYNCIO) | (1 << BIO_RW_UNPLUG); |
64b5a32e | 397 | res = nilfs_submit_seg_bio(wi, rw); |
64b5a32e RK |
398 | } |
399 | ||
64b5a32e | 400 | failed_bio: |
5f1586d0 | 401 | return res; |
64b5a32e RK |
402 | } |
403 | ||
404 | /** | |
405 | * nilfs_segbuf_wait - wait for completion of requested BIOs | |
406 | * @wi: nilfs_write_info | |
407 | * | |
408 | * Return Value: On Success, 0 is returned. On Error, one of the following | |
409 | * negative error code is returned. | |
410 | * | |
411 | * %-EIO - I/O error | |
412 | */ | |
413 | int nilfs_segbuf_wait(struct nilfs_segment_buffer *segbuf, | |
414 | struct nilfs_write_info *wi) | |
415 | { | |
416 | int err = 0; | |
417 | ||
418 | if (!wi->nbio) | |
419 | return 0; | |
420 | ||
421 | do { | |
422 | wait_for_completion(&wi->bio_event); | |
423 | } while (--wi->nbio > 0); | |
424 | ||
425 | if (unlikely(atomic_read(&wi->err) > 0)) { | |
426 | printk(KERN_ERR "NILFS: IO error writing segment\n"); | |
427 | err = -EIO; | |
428 | segbuf->sb_io_error = 1; | |
429 | } | |
430 | return err; | |
431 | } |