Commit | Line | Data |
---|---|---|
b3b94faa DT |
1 | /* |
2 | * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. | |
3a8a9a10 | 3 | * Copyright (C) 2004-2006 Red Hat, Inc. All rights reserved. |
b3b94faa DT |
4 | * |
5 | * This copyrighted material is made available to anyone wishing to use, | |
6 | * modify, copy, or redistribute it subject to the terms and conditions | |
e9fc2aa0 | 7 | * of the GNU General Public License version 2. |
b3b94faa DT |
8 | */ |
9 | ||
10 | #include <linux/sched.h> | |
11 | #include <linux/slab.h> | |
12 | #include <linux/spinlock.h> | |
13 | #include <linux/completion.h> | |
14 | #include <linux/buffer_head.h> | |
75ca61c1 | 15 | #include <linux/mempool.h> |
5c676f6d | 16 | #include <linux/gfs2_ondisk.h> |
c969f58c SW |
17 | #include <linux/bio.h> |
18 | #include <linux/fs.h> | |
7f63257d | 19 | #include <linux/list_sort.h> |
b3b94faa DT |
20 | |
21 | #include "gfs2.h" | |
5c676f6d | 22 | #include "incore.h" |
2332c443 | 23 | #include "inode.h" |
b3b94faa DT |
24 | #include "glock.h" |
25 | #include "log.h" | |
26 | #include "lops.h" | |
27 | #include "meta_io.h" | |
28 | #include "recovery.h" | |
29 | #include "rgrp.h" | |
30 | #include "trans.h" | |
5c676f6d | 31 | #include "util.h" |
63997775 | 32 | #include "trace_gfs2.h" |
b3b94faa | 33 | |
9b9107a5 SW |
34 | /** |
35 | * gfs2_pin - Pin a buffer in memory | |
36 | * @sdp: The superblock | |
37 | * @bh: The buffer to be pinned | |
38 | * | |
39 | * The log lock must be held when calling this function | |
40 | */ | |
767f433f | 41 | void gfs2_pin(struct gfs2_sbd *sdp, struct buffer_head *bh) |
9b9107a5 SW |
42 | { |
43 | struct gfs2_bufdata *bd; | |
44 | ||
29687a2a | 45 | BUG_ON(!current->journal_info); |
9b9107a5 SW |
46 | |
47 | clear_buffer_dirty(bh); | |
48 | if (test_set_buffer_pinned(bh)) | |
49 | gfs2_assert_withdraw(sdp, 0); | |
50 | if (!buffer_uptodate(bh)) | |
51 | gfs2_io_error_bh(sdp, bh); | |
52 | bd = bh->b_private; | |
53 | /* If this buffer is in the AIL and it has already been written | |
54 | * to in-place disk block, remove it from the AIL. | |
55 | */ | |
c618e87a | 56 | spin_lock(&sdp->sd_ail_lock); |
16ca9412 BM |
57 | if (bd->bd_tr) |
58 | list_move(&bd->bd_ail_st_list, &bd->bd_tr->tr_ail2_list); | |
c618e87a | 59 | spin_unlock(&sdp->sd_ail_lock); |
9b9107a5 | 60 | get_bh(bh); |
5e687eac | 61 | atomic_inc(&sdp->sd_log_pinned); |
63997775 | 62 | trace_gfs2_pin(bd, 1); |
9b9107a5 SW |
63 | } |
64 | ||
7c9ca621 BP |
65 | static bool buffer_is_rgrp(const struct gfs2_bufdata *bd) |
66 | { | |
67 | return bd->bd_gl->gl_name.ln_type == LM_TYPE_RGRP; | |
68 | } | |
69 | ||
70 | static void maybe_release_space(struct gfs2_bufdata *bd) | |
71 | { | |
72 | struct gfs2_glock *gl = bd->bd_gl; | |
15562c43 | 73 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
6f6597ba | 74 | struct gfs2_rgrpd *rgd = gfs2_glock2rgrp(gl); |
7c9ca621 BP |
75 | unsigned int index = bd->bd_bh->b_blocknr - gl->gl_name.ln_number; |
76 | struct gfs2_bitmap *bi = rgd->rd_bits + index; | |
77 | ||
5a7c6690 | 78 | if (bi->bi_clone == NULL) |
7c9ca621 BP |
79 | return; |
80 | if (sdp->sd_args.ar_discard) | |
66fc061b | 81 | gfs2_rgrp_send_discards(sdp, rgd->rd_data0, bd->bd_bh, bi, 1, NULL); |
7c9ca621 BP |
82 | memcpy(bi->bi_clone + bi->bi_offset, |
83 | bd->bd_bh->b_data + bi->bi_offset, bi->bi_len); | |
84 | clear_bit(GBF_FULL, &bi->bi_flags); | |
85 | rgd->rd_free_clone = rgd->rd_free; | |
5ea5050c | 86 | rgd->rd_extfail_pt = rgd->rd_free; |
7c9ca621 BP |
87 | } |
88 | ||
9b9107a5 SW |
89 | /** |
90 | * gfs2_unpin - Unpin a buffer | |
91 | * @sdp: the filesystem the buffer belongs to | |
92 | * @bh: The buffer to unpin | |
93 | * @ai: | |
29687a2a | 94 | * @flags: The inode dirty flags |
9b9107a5 SW |
95 | * |
96 | */ | |
97 | ||
98 | static void gfs2_unpin(struct gfs2_sbd *sdp, struct buffer_head *bh, | |
16ca9412 | 99 | struct gfs2_trans *tr) |
9b9107a5 SW |
100 | { |
101 | struct gfs2_bufdata *bd = bh->b_private; | |
102 | ||
29687a2a SW |
103 | BUG_ON(!buffer_uptodate(bh)); |
104 | BUG_ON(!buffer_pinned(bh)); | |
9b9107a5 SW |
105 | |
106 | lock_buffer(bh); | |
107 | mark_buffer_dirty(bh); | |
108 | clear_buffer_pinned(bh); | |
109 | ||
7c9ca621 BP |
110 | if (buffer_is_rgrp(bd)) |
111 | maybe_release_space(bd); | |
112 | ||
d6a079e8 | 113 | spin_lock(&sdp->sd_ail_lock); |
16ca9412 | 114 | if (bd->bd_tr) { |
9b9107a5 SW |
115 | list_del(&bd->bd_ail_st_list); |
116 | brelse(bh); | |
117 | } else { | |
118 | struct gfs2_glock *gl = bd->bd_gl; | |
119 | list_add(&bd->bd_ail_gl_list, &gl->gl_ail_list); | |
120 | atomic_inc(&gl->gl_ail_count); | |
121 | } | |
16ca9412 BM |
122 | bd->bd_tr = tr; |
123 | list_add(&bd->bd_ail_st_list, &tr->tr_ail1_list); | |
d6a079e8 DC |
124 | spin_unlock(&sdp->sd_ail_lock); |
125 | ||
29687a2a | 126 | clear_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags); |
63997775 | 127 | trace_gfs2_pin(bd, 0); |
9b9107a5 | 128 | unlock_buffer(bh); |
5e687eac | 129 | atomic_dec(&sdp->sd_log_pinned); |
9b9107a5 SW |
130 | } |
131 | ||
e8c92ed7 | 132 | static void gfs2_log_incr_head(struct gfs2_sbd *sdp) |
16615be1 | 133 | { |
e8c92ed7 SW |
134 | BUG_ON((sdp->sd_log_flush_head == sdp->sd_log_tail) && |
135 | (sdp->sd_log_flush_head != sdp->sd_log_head)); | |
136 | ||
722f6f62 | 137 | if (++sdp->sd_log_flush_head == sdp->sd_jdesc->jd_blocks) |
e8c92ed7 | 138 | sdp->sd_log_flush_head = 0; |
16615be1 SW |
139 | } |
140 | ||
e8c92ed7 | 141 | static u64 gfs2_log_bmap(struct gfs2_sbd *sdp) |
16615be1 | 142 | { |
e8c92ed7 SW |
143 | unsigned int lbn = sdp->sd_log_flush_head; |
144 | struct gfs2_journal_extent *je; | |
145 | u64 block; | |
146 | ||
b50f227b SW |
147 | list_for_each_entry(je, &sdp->sd_jdesc->extent_list, list) { |
148 | if ((lbn >= je->lblock) && (lbn < (je->lblock + je->blocks))) { | |
e8c92ed7 SW |
149 | block = je->dblock + lbn - je->lblock; |
150 | gfs2_log_incr_head(sdp); | |
151 | return block; | |
152 | } | |
153 | } | |
154 | ||
155 | return -1; | |
16615be1 SW |
156 | } |
157 | ||
e8c92ed7 SW |
158 | /** |
159 | * gfs2_end_log_write_bh - end log write of pagecache data with buffers | |
160 | * @sdp: The superblock | |
161 | * @bvec: The bio_vec | |
162 | * @error: The i/o status | |
163 | * | |
164 | * This finds the relavent buffers and unlocks then and sets the | |
165 | * error flag according to the status of the i/o request. This is | |
166 | * used when the log is writing data which has an in-place version | |
167 | * that is pinned in the pagecache. | |
168 | */ | |
169 | ||
170 | static void gfs2_end_log_write_bh(struct gfs2_sbd *sdp, struct bio_vec *bvec, | |
4e4cbee9 | 171 | blk_status_t error) |
16615be1 | 172 | { |
e8c92ed7 SW |
173 | struct buffer_head *bh, *next; |
174 | struct page *page = bvec->bv_page; | |
175 | unsigned size; | |
176 | ||
177 | bh = page_buffers(page); | |
178 | size = bvec->bv_len; | |
179 | while (bh_offset(bh) < bvec->bv_offset) | |
180 | bh = bh->b_this_page; | |
181 | do { | |
182 | if (error) | |
87354e5d | 183 | mark_buffer_write_io_error(bh); |
e8c92ed7 SW |
184 | unlock_buffer(bh); |
185 | next = bh->b_this_page; | |
186 | size -= bh->b_size; | |
187 | brelse(bh); | |
188 | bh = next; | |
189 | } while(bh && size); | |
16615be1 SW |
190 | } |
191 | ||
47ac5537 | 192 | /** |
e8c92ed7 SW |
193 | * gfs2_end_log_write - end of i/o to the log |
194 | * @bio: The bio | |
195 | * @error: Status of i/o request | |
196 | * | |
197 | * Each bio_vec contains either data from the pagecache or data | |
198 | * relating to the log itself. Here we iterate over the bio_vec | |
199 | * array, processing both kinds of data. | |
47ac5537 SW |
200 | * |
201 | */ | |
202 | ||
4246a0b6 | 203 | static void gfs2_end_log_write(struct bio *bio) |
47ac5537 | 204 | { |
e8c92ed7 SW |
205 | struct gfs2_sbd *sdp = bio->bi_private; |
206 | struct bio_vec *bvec; | |
207 | struct page *page; | |
208 | int i; | |
209 | ||
942b0cdd BP |
210 | if (bio->bi_status) { |
211 | fs_err(sdp, "Error %d writing to journal, jid=%u\n", | |
212 | bio->bi_status, sdp->sd_jdesc->jd_jid); | |
213 | wake_up(&sdp->sd_logd_waitq); | |
214 | } | |
e8c92ed7 | 215 | |
e97e548b | 216 | bio_for_each_segment_all(bvec, bio, i) { |
e8c92ed7 SW |
217 | page = bvec->bv_page; |
218 | if (page_has_buffers(page)) | |
4e4cbee9 | 219 | gfs2_end_log_write_bh(sdp, bvec, bio->bi_status); |
e8c92ed7 SW |
220 | else |
221 | mempool_free(page, gfs2_page_pool); | |
222 | } | |
47ac5537 | 223 | |
e8c92ed7 | 224 | bio_put(bio); |
47ac5537 SW |
225 | if (atomic_dec_and_test(&sdp->sd_log_in_flight)) |
226 | wake_up(&sdp->sd_log_flush_wait); | |
227 | } | |
228 | ||
229 | /** | |
e8c92ed7 SW |
230 | * gfs2_log_flush_bio - Submit any pending log bio |
231 | * @sdp: The superblock | |
e1b1afa6 | 232 | * @op: REQ_OP |
ef295ecf | 233 | * @op_flags: req_flag_bits |
47ac5537 | 234 | * |
e8c92ed7 SW |
235 | * Submit any pending part-built or full bio to the block device. If |
236 | * there is no pending bio, then this is a no-op. | |
47ac5537 SW |
237 | */ |
238 | ||
e1b1afa6 | 239 | void gfs2_log_flush_bio(struct gfs2_sbd *sdp, int op, int op_flags) |
47ac5537 | 240 | { |
e8c92ed7 SW |
241 | if (sdp->sd_log_bio) { |
242 | atomic_inc(&sdp->sd_log_in_flight); | |
e1b1afa6 | 243 | bio_set_op_attrs(sdp->sd_log_bio, op, op_flags); |
4e49ea4a | 244 | submit_bio(sdp->sd_log_bio); |
e8c92ed7 SW |
245 | sdp->sd_log_bio = NULL; |
246 | } | |
247 | } | |
47ac5537 | 248 | |
e8c92ed7 SW |
249 | /** |
250 | * gfs2_log_alloc_bio - Allocate a new bio for log writing | |
251 | * @sdp: The superblock | |
252 | * @blkno: The next device block number we want to write to | |
253 | * | |
254 | * This should never be called when there is a cached bio in the | |
255 | * super block. When it returns, there will be a cached bio in the | |
256 | * super block which will have as many bio_vecs as the device is | |
257 | * happy to handle. | |
258 | * | |
259 | * Returns: Newly allocated bio | |
260 | */ | |
261 | ||
262 | static struct bio *gfs2_log_alloc_bio(struct gfs2_sbd *sdp, u64 blkno) | |
263 | { | |
264 | struct super_block *sb = sdp->sd_vfs; | |
e8c92ed7 SW |
265 | struct bio *bio; |
266 | ||
267 | BUG_ON(sdp->sd_log_bio); | |
268 | ||
b54ffb73 | 269 | bio = bio_alloc(GFP_NOIO, BIO_MAX_PAGES); |
4f024f37 | 270 | bio->bi_iter.bi_sector = blkno * (sb->s_blocksize >> 9); |
74d46992 | 271 | bio_set_dev(bio, sb->s_bdev); |
e8c92ed7 SW |
272 | bio->bi_end_io = gfs2_end_log_write; |
273 | bio->bi_private = sdp; | |
274 | ||
275 | sdp->sd_log_bio = bio; | |
276 | ||
277 | return bio; | |
47ac5537 SW |
278 | } |
279 | ||
280 | /** | |
e8c92ed7 SW |
281 | * gfs2_log_get_bio - Get cached log bio, or allocate a new one |
282 | * @sdp: The superblock | |
283 | * @blkno: The device block number we want to write to | |
284 | * | |
285 | * If there is a cached bio, then if the next block number is sequential | |
286 | * with the previous one, return it, otherwise flush the bio to the | |
287 | * device. If there is not a cached bio, or we just flushed it, then | |
288 | * allocate a new one. | |
47ac5537 | 289 | * |
e8c92ed7 | 290 | * Returns: The bio to use for log writes |
47ac5537 SW |
291 | */ |
292 | ||
e8c92ed7 | 293 | static struct bio *gfs2_log_get_bio(struct gfs2_sbd *sdp, u64 blkno) |
47ac5537 | 294 | { |
e8c92ed7 SW |
295 | struct bio *bio = sdp->sd_log_bio; |
296 | u64 nblk; | |
297 | ||
298 | if (bio) { | |
f73a1c7d | 299 | nblk = bio_end_sector(bio); |
e8c92ed7 SW |
300 | nblk >>= sdp->sd_fsb2bb_shift; |
301 | if (blkno == nblk) | |
302 | return bio; | |
e1b1afa6 | 303 | gfs2_log_flush_bio(sdp, REQ_OP_WRITE, 0); |
e8c92ed7 SW |
304 | } |
305 | ||
306 | return gfs2_log_alloc_bio(sdp, blkno); | |
47ac5537 SW |
307 | } |
308 | ||
e8c92ed7 | 309 | |
47ac5537 | 310 | /** |
e8c92ed7 | 311 | * gfs2_log_write - write to log |
47ac5537 | 312 | * @sdp: the filesystem |
e8c92ed7 SW |
313 | * @page: the page to write |
314 | * @size: the size of the data to write | |
315 | * @offset: the offset within the page | |
47ac5537 | 316 | * |
e8c92ed7 SW |
317 | * Try and add the page segment to the current bio. If that fails, |
318 | * submit the current bio to the device and create a new one, and | |
319 | * then add the page segment to that. | |
47ac5537 SW |
320 | */ |
321 | ||
e8c92ed7 SW |
322 | static void gfs2_log_write(struct gfs2_sbd *sdp, struct page *page, |
323 | unsigned size, unsigned offset) | |
47ac5537 | 324 | { |
e8c92ed7 SW |
325 | u64 blkno = gfs2_log_bmap(sdp); |
326 | struct bio *bio; | |
327 | int ret; | |
328 | ||
329 | bio = gfs2_log_get_bio(sdp, blkno); | |
330 | ret = bio_add_page(bio, page, size, offset); | |
331 | if (ret == 0) { | |
e1b1afa6 | 332 | gfs2_log_flush_bio(sdp, REQ_OP_WRITE, 0); |
e8c92ed7 SW |
333 | bio = gfs2_log_alloc_bio(sdp, blkno); |
334 | ret = bio_add_page(bio, page, size, offset); | |
335 | WARN_ON(ret == 0); | |
336 | } | |
337 | } | |
47ac5537 | 338 | |
e8c92ed7 SW |
339 | /** |
340 | * gfs2_log_write_bh - write a buffer's content to the log | |
341 | * @sdp: The super block | |
342 | * @bh: The buffer pointing to the in-place location | |
343 | * | |
344 | * This writes the content of the buffer to the next available location | |
345 | * in the log. The buffer will be unlocked once the i/o to the log has | |
346 | * completed. | |
347 | */ | |
348 | ||
349 | static void gfs2_log_write_bh(struct gfs2_sbd *sdp, struct buffer_head *bh) | |
350 | { | |
351 | gfs2_log_write(sdp, bh->b_page, bh->b_size, bh_offset(bh)); | |
352 | } | |
47ac5537 | 353 | |
e8c92ed7 SW |
354 | /** |
355 | * gfs2_log_write_page - write one block stored in a page, into the log | |
356 | * @sdp: The superblock | |
357 | * @page: The struct page | |
358 | * | |
359 | * This writes the first block-sized part of the page into the log. Note | |
360 | * that the page must have been allocated from the gfs2_page_pool mempool | |
361 | * and that after this has been called, ownership has been transferred and | |
362 | * the page may be freed at any time. | |
363 | */ | |
47ac5537 | 364 | |
e8c92ed7 SW |
365 | void gfs2_log_write_page(struct gfs2_sbd *sdp, struct page *page) |
366 | { | |
367 | struct super_block *sb = sdp->sd_vfs; | |
368 | gfs2_log_write(sdp, page, sb->s_blocksize, 0); | |
47ac5537 | 369 | } |
16615be1 | 370 | |
dad30e90 SW |
371 | static struct page *gfs2_get_log_desc(struct gfs2_sbd *sdp, u32 ld_type, |
372 | u32 ld_length, u32 ld_data1) | |
16615be1 | 373 | { |
144a4c2f | 374 | struct page *page = mempool_alloc(gfs2_page_pool, GFP_NOIO); |
e8c92ed7 SW |
375 | struct gfs2_log_descriptor *ld = page_address(page); |
376 | clear_page(ld); | |
16615be1 SW |
377 | ld->ld_header.mh_magic = cpu_to_be32(GFS2_MAGIC); |
378 | ld->ld_header.mh_type = cpu_to_be32(GFS2_METATYPE_LD); | |
379 | ld->ld_header.mh_format = cpu_to_be32(GFS2_FORMAT_LD); | |
380 | ld->ld_type = cpu_to_be32(ld_type); | |
dad30e90 SW |
381 | ld->ld_length = cpu_to_be32(ld_length); |
382 | ld->ld_data1 = cpu_to_be32(ld_data1); | |
16615be1 | 383 | ld->ld_data2 = 0; |
e8c92ed7 | 384 | return page; |
16615be1 SW |
385 | } |
386 | ||
dad30e90 SW |
387 | static void gfs2_check_magic(struct buffer_head *bh) |
388 | { | |
389 | void *kaddr; | |
390 | __be32 *ptr; | |
391 | ||
392 | clear_buffer_escaped(bh); | |
393 | kaddr = kmap_atomic(bh->b_page); | |
394 | ptr = kaddr + bh_offset(bh); | |
395 | if (*ptr == cpu_to_be32(GFS2_MAGIC)) | |
396 | set_buffer_escaped(bh); | |
397 | kunmap_atomic(kaddr); | |
398 | } | |
399 | ||
7f63257d BM |
400 | static int blocknr_cmp(void *priv, struct list_head *a, struct list_head *b) |
401 | { | |
402 | struct gfs2_bufdata *bda, *bdb; | |
403 | ||
404 | bda = list_entry(a, struct gfs2_bufdata, bd_list); | |
405 | bdb = list_entry(b, struct gfs2_bufdata, bd_list); | |
406 | ||
407 | if (bda->bd_bh->b_blocknr < bdb->bd_bh->b_blocknr) | |
408 | return -1; | |
409 | if (bda->bd_bh->b_blocknr > bdb->bd_bh->b_blocknr) | |
410 | return 1; | |
411 | return 0; | |
412 | } | |
413 | ||
dad30e90 SW |
414 | static void gfs2_before_commit(struct gfs2_sbd *sdp, unsigned int limit, |
415 | unsigned int total, struct list_head *blist, | |
416 | bool is_databuf) | |
b3b94faa | 417 | { |
b3b94faa DT |
418 | struct gfs2_log_descriptor *ld; |
419 | struct gfs2_bufdata *bd1 = NULL, *bd2; | |
e8c92ed7 | 420 | struct page *page; |
b3b94faa DT |
421 | unsigned int num; |
422 | unsigned n; | |
423 | __be64 *ptr; | |
424 | ||
905d2aef | 425 | gfs2_log_lock(sdp); |
7f63257d | 426 | list_sort(NULL, blist, blocknr_cmp); |
c0752aa7 | 427 | bd1 = bd2 = list_prepare_entry(bd1, blist, bd_list); |
b3b94faa DT |
428 | while(total) { |
429 | num = total; | |
430 | if (total > limit) | |
431 | num = limit; | |
905d2aef | 432 | gfs2_log_unlock(sdp); |
4a586812 BP |
433 | page = gfs2_get_log_desc(sdp, |
434 | is_databuf ? GFS2_LOG_DESC_JDATA : | |
435 | GFS2_LOG_DESC_METADATA, num + 1, num); | |
e8c92ed7 | 436 | ld = page_address(page); |
905d2aef | 437 | gfs2_log_lock(sdp); |
e8c92ed7 | 438 | ptr = (__be64 *)(ld + 1); |
b3b94faa DT |
439 | |
440 | n = 0; | |
c0752aa7 | 441 | list_for_each_entry_continue(bd1, blist, bd_list) { |
b3b94faa | 442 | *ptr++ = cpu_to_be64(bd1->bd_bh->b_blocknr); |
dad30e90 SW |
443 | if (is_databuf) { |
444 | gfs2_check_magic(bd1->bd_bh); | |
445 | *ptr++ = cpu_to_be64(buffer_escaped(bd1->bd_bh) ? 1 : 0); | |
446 | } | |
b3b94faa DT |
447 | if (++n >= num) |
448 | break; | |
449 | } | |
450 | ||
905d2aef | 451 | gfs2_log_unlock(sdp); |
e8c92ed7 | 452 | gfs2_log_write_page(sdp, page); |
905d2aef | 453 | gfs2_log_lock(sdp); |
b3b94faa DT |
454 | |
455 | n = 0; | |
c0752aa7 | 456 | list_for_each_entry_continue(bd2, blist, bd_list) { |
16615be1 | 457 | get_bh(bd2->bd_bh); |
905d2aef | 458 | gfs2_log_unlock(sdp); |
16615be1 | 459 | lock_buffer(bd2->bd_bh); |
dad30e90 SW |
460 | |
461 | if (buffer_escaped(bd2->bd_bh)) { | |
462 | void *kaddr; | |
463 | page = mempool_alloc(gfs2_page_pool, GFP_NOIO); | |
464 | ptr = page_address(page); | |
465 | kaddr = kmap_atomic(bd2->bd_bh->b_page); | |
466 | memcpy(ptr, kaddr + bh_offset(bd2->bd_bh), | |
467 | bd2->bd_bh->b_size); | |
468 | kunmap_atomic(kaddr); | |
469 | *(__be32 *)ptr = 0; | |
470 | clear_buffer_escaped(bd2->bd_bh); | |
471 | unlock_buffer(bd2->bd_bh); | |
472 | brelse(bd2->bd_bh); | |
473 | gfs2_log_write_page(sdp, page); | |
474 | } else { | |
475 | gfs2_log_write_bh(sdp, bd2->bd_bh); | |
476 | } | |
905d2aef | 477 | gfs2_log_lock(sdp); |
b3b94faa DT |
478 | if (++n >= num) |
479 | break; | |
480 | } | |
481 | ||
905d2aef | 482 | BUG_ON(total < num); |
b3b94faa DT |
483 | total -= num; |
484 | } | |
905d2aef | 485 | gfs2_log_unlock(sdp); |
b3b94faa DT |
486 | } |
487 | ||
d69a3c65 | 488 | static void buf_lo_before_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr) |
dad30e90 SW |
489 | { |
490 | unsigned int limit = buf_limit(sdp); /* 503 for 4k blocks */ | |
022ef4fe | 491 | unsigned int nbuf; |
d69a3c65 SW |
492 | if (tr == NULL) |
493 | return; | |
022ef4fe SW |
494 | nbuf = tr->tr_num_buf_new - tr->tr_num_buf_rm; |
495 | gfs2_before_commit(sdp, limit, nbuf, &tr->tr_buf, 0); | |
dad30e90 SW |
496 | } |
497 | ||
16ca9412 | 498 | static void buf_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr) |
b3b94faa | 499 | { |
d69a3c65 | 500 | struct list_head *head; |
b3b94faa DT |
501 | struct gfs2_bufdata *bd; |
502 | ||
d69a3c65 | 503 | if (tr == NULL) |
16ca9412 | 504 | return; |
16ca9412 | 505 | |
d69a3c65 | 506 | head = &tr->tr_buf; |
b3b94faa | 507 | while (!list_empty(head)) { |
c0752aa7 BP |
508 | bd = list_entry(head->next, struct gfs2_bufdata, bd_list); |
509 | list_del_init(&bd->bd_list); | |
16ca9412 | 510 | gfs2_unpin(sdp, bd->bd_bh, tr); |
b3b94faa | 511 | } |
b3b94faa DT |
512 | } |
513 | ||
514 | static void buf_lo_before_scan(struct gfs2_jdesc *jd, | |
55167622 | 515 | struct gfs2_log_header_host *head, int pass) |
b3b94faa | 516 | { |
b3b94faa DT |
517 | if (pass != 0) |
518 | return; | |
519 | ||
a17d758b BP |
520 | jd->jd_found_blocks = 0; |
521 | jd->jd_replayed_blocks = 0; | |
b3b94faa DT |
522 | } |
523 | ||
524 | static int buf_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start, | |
525 | struct gfs2_log_descriptor *ld, __be64 *ptr, | |
526 | int pass) | |
527 | { | |
feaa7bba SW |
528 | struct gfs2_inode *ip = GFS2_I(jd->jd_inode); |
529 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); | |
5c676f6d | 530 | struct gfs2_glock *gl = ip->i_gl; |
b3b94faa DT |
531 | unsigned int blks = be32_to_cpu(ld->ld_data1); |
532 | struct buffer_head *bh_log, *bh_ip; | |
cd915493 | 533 | u64 blkno; |
b3b94faa DT |
534 | int error = 0; |
535 | ||
536 | if (pass != 1 || be32_to_cpu(ld->ld_type) != GFS2_LOG_DESC_METADATA) | |
537 | return 0; | |
538 | ||
e1cb6be9 | 539 | gfs2_replay_incr_blk(jd, &start); |
b3b94faa | 540 | |
e1cb6be9 | 541 | for (; blks; gfs2_replay_incr_blk(jd, &start), blks--) { |
b3b94faa DT |
542 | blkno = be64_to_cpu(*ptr++); |
543 | ||
a17d758b | 544 | jd->jd_found_blocks++; |
b3b94faa | 545 | |
a17d758b | 546 | if (gfs2_revoke_check(jd, blkno, start)) |
b3b94faa DT |
547 | continue; |
548 | ||
549 | error = gfs2_replay_read_block(jd, start, &bh_log); | |
82ffa516 SW |
550 | if (error) |
551 | return error; | |
b3b94faa DT |
552 | |
553 | bh_ip = gfs2_meta_new(gl, blkno); | |
554 | memcpy(bh_ip->b_data, bh_log->b_data, bh_log->b_size); | |
555 | ||
556 | if (gfs2_meta_check(sdp, bh_ip)) | |
557 | error = -EIO; | |
558 | else | |
559 | mark_buffer_dirty(bh_ip); | |
560 | ||
561 | brelse(bh_log); | |
562 | brelse(bh_ip); | |
563 | ||
564 | if (error) | |
565 | break; | |
566 | ||
a17d758b | 567 | jd->jd_replayed_blocks++; |
b3b94faa DT |
568 | } |
569 | ||
570 | return error; | |
571 | } | |
572 | ||
7c0ef28a SW |
573 | /** |
574 | * gfs2_meta_sync - Sync all buffers associated with a glock | |
575 | * @gl: The glock | |
576 | * | |
577 | */ | |
578 | ||
579 | static void gfs2_meta_sync(struct gfs2_glock *gl) | |
580 | { | |
581 | struct address_space *mapping = gfs2_glock2aspace(gl); | |
15562c43 | 582 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
7c0ef28a SW |
583 | int error; |
584 | ||
70d4ee94 SW |
585 | if (mapping == NULL) |
586 | mapping = &sdp->sd_aspace; | |
587 | ||
7c0ef28a SW |
588 | filemap_fdatawrite(mapping); |
589 | error = filemap_fdatawait(mapping); | |
590 | ||
591 | if (error) | |
15562c43 | 592 | gfs2_io_error(gl->gl_name.ln_sbd); |
7c0ef28a SW |
593 | } |
594 | ||
b3b94faa DT |
595 | static void buf_lo_after_scan(struct gfs2_jdesc *jd, int error, int pass) |
596 | { | |
feaa7bba SW |
597 | struct gfs2_inode *ip = GFS2_I(jd->jd_inode); |
598 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); | |
b3b94faa DT |
599 | |
600 | if (error) { | |
7276b3b0 | 601 | gfs2_meta_sync(ip->i_gl); |
b3b94faa DT |
602 | return; |
603 | } | |
604 | if (pass != 1) | |
605 | return; | |
606 | ||
7276b3b0 | 607 | gfs2_meta_sync(ip->i_gl); |
b3b94faa DT |
608 | |
609 | fs_info(sdp, "jid=%u: Replayed %u of %u blocks\n", | |
a17d758b | 610 | jd->jd_jid, jd->jd_replayed_blocks, jd->jd_found_blocks); |
b3b94faa DT |
611 | } |
612 | ||
d69a3c65 | 613 | static void revoke_lo_before_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr) |
b3b94faa | 614 | { |
b3b94faa | 615 | struct gfs2_meta_header *mh; |
b3b94faa DT |
616 | unsigned int offset; |
617 | struct list_head *head = &sdp->sd_log_le_revoke; | |
82e86087 | 618 | struct gfs2_bufdata *bd; |
e8c92ed7 | 619 | struct page *page; |
dad30e90 | 620 | unsigned int length; |
b3b94faa | 621 | |
5d054964 | 622 | gfs2_write_revokes(sdp); |
b3b94faa DT |
623 | if (!sdp->sd_log_num_revoke) |
624 | return; | |
625 | ||
dad30e90 SW |
626 | length = gfs2_struct2blk(sdp, sdp->sd_log_num_revoke, sizeof(u64)); |
627 | page = gfs2_get_log_desc(sdp, GFS2_LOG_DESC_REVOKE, length, sdp->sd_log_num_revoke); | |
b3b94faa DT |
628 | offset = sizeof(struct gfs2_log_descriptor); |
629 | ||
c0752aa7 | 630 | list_for_each_entry(bd, head, bd_list) { |
b3b94faa DT |
631 | sdp->sd_log_num_revoke--; |
632 | ||
cd915493 | 633 | if (offset + sizeof(u64) > sdp->sd_sb.sb_bsize) { |
b3b94faa | 634 | |
e8c92ed7 SW |
635 | gfs2_log_write_page(sdp, page); |
636 | page = mempool_alloc(gfs2_page_pool, GFP_NOIO); | |
637 | mh = page_address(page); | |
638 | clear_page(mh); | |
b3b94faa | 639 | mh->mh_magic = cpu_to_be32(GFS2_MAGIC); |
e3167ded SW |
640 | mh->mh_type = cpu_to_be32(GFS2_METATYPE_LB); |
641 | mh->mh_format = cpu_to_be32(GFS2_FORMAT_LB); | |
b3b94faa DT |
642 | offset = sizeof(struct gfs2_meta_header); |
643 | } | |
644 | ||
e8c92ed7 | 645 | *(__be64 *)(page_address(page) + offset) = cpu_to_be64(bd->bd_blkno); |
cd915493 | 646 | offset += sizeof(u64); |
b3b94faa DT |
647 | } |
648 | gfs2_assert_withdraw(sdp, !sdp->sd_log_num_revoke); | |
649 | ||
e8c92ed7 | 650 | gfs2_log_write_page(sdp, page); |
b3b94faa DT |
651 | } |
652 | ||
16ca9412 | 653 | static void revoke_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr) |
f42ab085 SW |
654 | { |
655 | struct list_head *head = &sdp->sd_log_le_revoke; | |
656 | struct gfs2_bufdata *bd; | |
657 | struct gfs2_glock *gl; | |
658 | ||
659 | while (!list_empty(head)) { | |
c0752aa7 BP |
660 | bd = list_entry(head->next, struct gfs2_bufdata, bd_list); |
661 | list_del_init(&bd->bd_list); | |
f42ab085 SW |
662 | gl = bd->bd_gl; |
663 | atomic_dec(&gl->gl_revokes); | |
664 | clear_bit(GLF_LFLUSH, &gl->gl_flags); | |
665 | kmem_cache_free(gfs2_bufdata_cachep, bd); | |
666 | } | |
667 | } | |
668 | ||
b3b94faa | 669 | static void revoke_lo_before_scan(struct gfs2_jdesc *jd, |
55167622 | 670 | struct gfs2_log_header_host *head, int pass) |
b3b94faa | 671 | { |
b3b94faa DT |
672 | if (pass != 0) |
673 | return; | |
674 | ||
a17d758b BP |
675 | jd->jd_found_revokes = 0; |
676 | jd->jd_replay_tail = head->lh_tail; | |
b3b94faa DT |
677 | } |
678 | ||
679 | static int revoke_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start, | |
680 | struct gfs2_log_descriptor *ld, __be64 *ptr, | |
681 | int pass) | |
682 | { | |
feaa7bba | 683 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); |
b3b94faa DT |
684 | unsigned int blks = be32_to_cpu(ld->ld_length); |
685 | unsigned int revokes = be32_to_cpu(ld->ld_data1); | |
686 | struct buffer_head *bh; | |
687 | unsigned int offset; | |
cd915493 | 688 | u64 blkno; |
b3b94faa DT |
689 | int first = 1; |
690 | int error; | |
691 | ||
692 | if (pass != 0 || be32_to_cpu(ld->ld_type) != GFS2_LOG_DESC_REVOKE) | |
693 | return 0; | |
694 | ||
695 | offset = sizeof(struct gfs2_log_descriptor); | |
696 | ||
e1cb6be9 | 697 | for (; blks; gfs2_replay_incr_blk(jd, &start), blks--) { |
b3b94faa DT |
698 | error = gfs2_replay_read_block(jd, start, &bh); |
699 | if (error) | |
700 | return error; | |
701 | ||
702 | if (!first) | |
703 | gfs2_metatype_check(sdp, bh, GFS2_METATYPE_LB); | |
704 | ||
cd915493 | 705 | while (offset + sizeof(u64) <= sdp->sd_sb.sb_bsize) { |
b3b94faa DT |
706 | blkno = be64_to_cpu(*(__be64 *)(bh->b_data + offset)); |
707 | ||
a17d758b | 708 | error = gfs2_revoke_add(jd, blkno, start); |
3ad62e87 BP |
709 | if (error < 0) { |
710 | brelse(bh); | |
b3b94faa | 711 | return error; |
3ad62e87 | 712 | } |
b3b94faa | 713 | else if (error) |
a17d758b | 714 | jd->jd_found_revokes++; |
b3b94faa DT |
715 | |
716 | if (!--revokes) | |
717 | break; | |
cd915493 | 718 | offset += sizeof(u64); |
b3b94faa DT |
719 | } |
720 | ||
721 | brelse(bh); | |
722 | offset = sizeof(struct gfs2_meta_header); | |
723 | first = 0; | |
724 | } | |
725 | ||
726 | return 0; | |
727 | } | |
728 | ||
729 | static void revoke_lo_after_scan(struct gfs2_jdesc *jd, int error, int pass) | |
730 | { | |
feaa7bba | 731 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); |
b3b94faa DT |
732 | |
733 | if (error) { | |
a17d758b | 734 | gfs2_revoke_clean(jd); |
b3b94faa DT |
735 | return; |
736 | } | |
737 | if (pass != 1) | |
738 | return; | |
739 | ||
740 | fs_info(sdp, "jid=%u: Found %u revoke tags\n", | |
a17d758b | 741 | jd->jd_jid, jd->jd_found_revokes); |
b3b94faa | 742 | |
a17d758b | 743 | gfs2_revoke_clean(jd); |
b3b94faa DT |
744 | } |
745 | ||
16615be1 SW |
746 | /** |
747 | * databuf_lo_before_commit - Scan the data buffers, writing as we go | |
748 | * | |
749 | */ | |
750 | ||
d69a3c65 | 751 | static void databuf_lo_before_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr) |
16615be1 | 752 | { |
022ef4fe SW |
753 | unsigned int limit = databuf_limit(sdp); |
754 | unsigned int nbuf; | |
d69a3c65 SW |
755 | if (tr == NULL) |
756 | return; | |
022ef4fe SW |
757 | nbuf = tr->tr_num_databuf_new - tr->tr_num_databuf_rm; |
758 | gfs2_before_commit(sdp, limit, nbuf, &tr->tr_databuf, 1); | |
18ec7d5c SW |
759 | } |
760 | ||
761 | static int databuf_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start, | |
762 | struct gfs2_log_descriptor *ld, | |
763 | __be64 *ptr, int pass) | |
764 | { | |
feaa7bba | 765 | struct gfs2_inode *ip = GFS2_I(jd->jd_inode); |
5c676f6d | 766 | struct gfs2_glock *gl = ip->i_gl; |
18ec7d5c SW |
767 | unsigned int blks = be32_to_cpu(ld->ld_data1); |
768 | struct buffer_head *bh_log, *bh_ip; | |
cd915493 SW |
769 | u64 blkno; |
770 | u64 esc; | |
18ec7d5c SW |
771 | int error = 0; |
772 | ||
773 | if (pass != 1 || be32_to_cpu(ld->ld_type) != GFS2_LOG_DESC_JDATA) | |
774 | return 0; | |
775 | ||
e1cb6be9 BP |
776 | gfs2_replay_incr_blk(jd, &start); |
777 | for (; blks; gfs2_replay_incr_blk(jd, &start), blks--) { | |
18ec7d5c SW |
778 | blkno = be64_to_cpu(*ptr++); |
779 | esc = be64_to_cpu(*ptr++); | |
780 | ||
a17d758b | 781 | jd->jd_found_blocks++; |
18ec7d5c | 782 | |
a17d758b | 783 | if (gfs2_revoke_check(jd, blkno, start)) |
18ec7d5c SW |
784 | continue; |
785 | ||
786 | error = gfs2_replay_read_block(jd, start, &bh_log); | |
787 | if (error) | |
788 | return error; | |
789 | ||
790 | bh_ip = gfs2_meta_new(gl, blkno); | |
791 | memcpy(bh_ip->b_data, bh_log->b_data, bh_log->b_size); | |
792 | ||
793 | /* Unescape */ | |
794 | if (esc) { | |
795 | __be32 *eptr = (__be32 *)bh_ip->b_data; | |
796 | *eptr = cpu_to_be32(GFS2_MAGIC); | |
797 | } | |
798 | mark_buffer_dirty(bh_ip); | |
799 | ||
800 | brelse(bh_log); | |
801 | brelse(bh_ip); | |
18ec7d5c | 802 | |
a17d758b | 803 | jd->jd_replayed_blocks++; |
18ec7d5c SW |
804 | } |
805 | ||
806 | return error; | |
807 | } | |
808 | ||
809 | /* FIXME: sort out accounting for log blocks etc. */ | |
810 | ||
811 | static void databuf_lo_after_scan(struct gfs2_jdesc *jd, int error, int pass) | |
812 | { | |
feaa7bba SW |
813 | struct gfs2_inode *ip = GFS2_I(jd->jd_inode); |
814 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); | |
18ec7d5c SW |
815 | |
816 | if (error) { | |
7276b3b0 | 817 | gfs2_meta_sync(ip->i_gl); |
18ec7d5c SW |
818 | return; |
819 | } | |
820 | if (pass != 1) | |
821 | return; | |
822 | ||
823 | /* data sync? */ | |
7276b3b0 | 824 | gfs2_meta_sync(ip->i_gl); |
18ec7d5c SW |
825 | |
826 | fs_info(sdp, "jid=%u: Replayed %u of %u data blocks\n", | |
a17d758b | 827 | jd->jd_jid, jd->jd_replayed_blocks, jd->jd_found_blocks); |
18ec7d5c SW |
828 | } |
829 | ||
16ca9412 | 830 | static void databuf_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr) |
18ec7d5c | 831 | { |
d69a3c65 | 832 | struct list_head *head; |
18ec7d5c SW |
833 | struct gfs2_bufdata *bd; |
834 | ||
d69a3c65 | 835 | if (tr == NULL) |
16ca9412 | 836 | return; |
16ca9412 | 837 | |
d69a3c65 | 838 | head = &tr->tr_databuf; |
18ec7d5c | 839 | while (!list_empty(head)) { |
c0752aa7 BP |
840 | bd = list_entry(head->next, struct gfs2_bufdata, bd_list); |
841 | list_del_init(&bd->bd_list); | |
16ca9412 | 842 | gfs2_unpin(sdp, bd->bd_bh, tr); |
18ec7d5c | 843 | } |
b3b94faa DT |
844 | } |
845 | ||
18ec7d5c | 846 | |
b09e593d | 847 | const struct gfs2_log_operations gfs2_buf_lops = { |
b3b94faa DT |
848 | .lo_before_commit = buf_lo_before_commit, |
849 | .lo_after_commit = buf_lo_after_commit, | |
850 | .lo_before_scan = buf_lo_before_scan, | |
851 | .lo_scan_elements = buf_lo_scan_elements, | |
852 | .lo_after_scan = buf_lo_after_scan, | |
ea67eedb | 853 | .lo_name = "buf", |
b3b94faa DT |
854 | }; |
855 | ||
b09e593d | 856 | const struct gfs2_log_operations gfs2_revoke_lops = { |
b3b94faa | 857 | .lo_before_commit = revoke_lo_before_commit, |
f42ab085 | 858 | .lo_after_commit = revoke_lo_after_commit, |
b3b94faa DT |
859 | .lo_before_scan = revoke_lo_before_scan, |
860 | .lo_scan_elements = revoke_lo_scan_elements, | |
861 | .lo_after_scan = revoke_lo_after_scan, | |
ea67eedb | 862 | .lo_name = "revoke", |
b3b94faa DT |
863 | }; |
864 | ||
b09e593d | 865 | const struct gfs2_log_operations gfs2_databuf_lops = { |
b3b94faa | 866 | .lo_before_commit = databuf_lo_before_commit, |
18ec7d5c SW |
867 | .lo_after_commit = databuf_lo_after_commit, |
868 | .lo_scan_elements = databuf_lo_scan_elements, | |
869 | .lo_after_scan = databuf_lo_after_scan, | |
ea67eedb | 870 | .lo_name = "databuf", |
b3b94faa DT |
871 | }; |
872 | ||
b09e593d | 873 | const struct gfs2_log_operations *gfs2_log_ops[] = { |
16615be1 | 874 | &gfs2_databuf_lops, |
b3b94faa | 875 | &gfs2_buf_lops, |
16615be1 | 876 | &gfs2_revoke_lops, |
ea67eedb | 877 | NULL, |
b3b94faa DT |
878 | }; |
879 |