1 // SPDX-License-Identifier: GPL-2.0-only <<
2 /* 1 /*
3 * Copyright (C) Sistina Software, Inc. 1997- 2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
4 * Copyright (C) 2004-2006 Red Hat, Inc. All 3 * Copyright (C) 2004-2006 Red Hat, Inc. All rights reserved.
>> 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
>> 7 * of the GNU General Public License version 2.
5 */ 8 */
6 9
7 #include <linux/sched.h> 10 #include <linux/sched.h>
8 #include <linux/slab.h> 11 #include <linux/slab.h>
9 #include <linux/spinlock.h> 12 #include <linux/spinlock.h>
10 #include <linux/completion.h> 13 #include <linux/completion.h>
11 #include <linux/buffer_head.h> 14 #include <linux/buffer_head.h>
12 #include <linux/mempool.h> <<
13 #include <linux/gfs2_ondisk.h> 15 #include <linux/gfs2_ondisk.h>
14 #include <linux/bio.h> 16 #include <linux/bio.h>
15 #include <linux/fs.h> 17 #include <linux/fs.h>
16 #include <linux/list_sort.h> <<
17 #include <linux/blkdev.h> <<
18 18
19 #include "bmap.h" <<
20 #include "dir.h" <<
21 #include "gfs2.h" 19 #include "gfs2.h"
22 #include "incore.h" 20 #include "incore.h"
23 #include "inode.h" 21 #include "inode.h"
24 #include "glock.h" 22 #include "glock.h"
25 #include "glops.h" <<
26 #include "log.h" 23 #include "log.h"
27 #include "lops.h" 24 #include "lops.h"
28 #include "meta_io.h" 25 #include "meta_io.h"
29 #include "recovery.h" 26 #include "recovery.h"
30 #include "rgrp.h" 27 #include "rgrp.h"
31 #include "trans.h" 28 #include "trans.h"
32 #include "util.h" 29 #include "util.h"
33 #include "trace_gfs2.h" 30 #include "trace_gfs2.h"
34 31
35 /** 32 /**
36 * gfs2_pin - Pin a buffer in memory 33 * gfs2_pin - Pin a buffer in memory
37 * @sdp: The superblock 34 * @sdp: The superblock
38 * @bh: The buffer to be pinned 35 * @bh: The buffer to be pinned
39 * 36 *
40 * The log lock must be held when calling this 37 * The log lock must be held when calling this function
41 */ 38 */
42 void gfs2_pin(struct gfs2_sbd *sdp, struct buf !! 39 static void gfs2_pin(struct gfs2_sbd *sdp, struct buffer_head *bh)
43 { 40 {
44 struct gfs2_bufdata *bd; 41 struct gfs2_bufdata *bd;
45 42
46 BUG_ON(!current->journal_info); !! 43 gfs2_assert_withdraw(sdp, test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags));
47 44
48 clear_buffer_dirty(bh); 45 clear_buffer_dirty(bh);
49 if (test_set_buffer_pinned(bh)) 46 if (test_set_buffer_pinned(bh))
50 gfs2_assert_withdraw(sdp, 0); 47 gfs2_assert_withdraw(sdp, 0);
51 if (!buffer_uptodate(bh)) 48 if (!buffer_uptodate(bh))
52 gfs2_io_error_bh_wd(sdp, bh); !! 49 gfs2_io_error_bh(sdp, bh);
53 bd = bh->b_private; 50 bd = bh->b_private;
54 /* If this buffer is in the AIL and it 51 /* If this buffer is in the AIL and it has already been written
55 * to in-place disk block, remove it f 52 * to in-place disk block, remove it from the AIL.
56 */ 53 */
57 spin_lock(&sdp->sd_ail_lock); !! 54 if (bd->bd_ail)
58 if (bd->bd_tr) !! 55 list_move(&bd->bd_ail_st_list, &bd->bd_ail->ai_ail2_list);
59 list_move(&bd->bd_ail_st_list, <<
60 spin_unlock(&sdp->sd_ail_lock); <<
61 get_bh(bh); 56 get_bh(bh);
62 atomic_inc(&sdp->sd_log_pinned); <<
63 trace_gfs2_pin(bd, 1); 57 trace_gfs2_pin(bd, 1);
64 } 58 }
65 59
66 static bool buffer_is_rgrp(const struct gfs2_b <<
67 { <<
68 return bd->bd_gl->gl_name.ln_type == L <<
69 } <<
70 <<
71 static void maybe_release_space(struct gfs2_bu <<
72 { <<
73 struct gfs2_glock *gl = bd->bd_gl; <<
74 struct gfs2_sbd *sdp = gl->gl_name.ln_ <<
75 struct gfs2_rgrpd *rgd = gfs2_glock2rg <<
76 unsigned int index = bd->bd_bh->b_bloc <<
77 struct gfs2_bitmap *bi = rgd->rd_bits <<
78 <<
79 rgrp_lock_local(rgd); <<
80 if (bi->bi_clone == NULL) <<
81 goto out; <<
82 if (sdp->sd_args.ar_discard) <<
83 gfs2_rgrp_send_discards(sdp, r <<
84 memcpy(bi->bi_clone + bi->bi_offset, <<
85 bd->bd_bh->b_data + bi->bi_offs <<
86 clear_bit(GBF_FULL, &bi->bi_flags); <<
87 rgd->rd_free_clone = rgd->rd_free; <<
88 BUG_ON(rgd->rd_free_clone < rgd->rd_re <<
89 rgd->rd_extfail_pt = rgd->rd_free; <<
90 <<
91 out: <<
92 rgrp_unlock_local(rgd); <<
93 } <<
94 <<
95 /** 60 /**
96 * gfs2_unpin - Unpin a buffer 61 * gfs2_unpin - Unpin a buffer
97 * @sdp: the filesystem the buffer belongs to 62 * @sdp: the filesystem the buffer belongs to
98 * @bh: The buffer to unpin 63 * @bh: The buffer to unpin
99 * @tr: The system transaction being flushed !! 64 * @ai:
>> 65 *
100 */ 66 */
101 67
102 static void gfs2_unpin(struct gfs2_sbd *sdp, s 68 static void gfs2_unpin(struct gfs2_sbd *sdp, struct buffer_head *bh,
103 struct gfs2_trans *tr) !! 69 struct gfs2_ail *ai)
104 { 70 {
105 struct gfs2_bufdata *bd = bh->b_privat 71 struct gfs2_bufdata *bd = bh->b_private;
106 72
107 BUG_ON(!buffer_uptodate(bh)); !! 73 gfs2_assert_withdraw(sdp, buffer_uptodate(bh));
108 BUG_ON(!buffer_pinned(bh)); !! 74
>> 75 if (!buffer_pinned(bh))
>> 76 gfs2_assert_withdraw(sdp, 0);
109 77
110 lock_buffer(bh); 78 lock_buffer(bh);
111 mark_buffer_dirty(bh); 79 mark_buffer_dirty(bh);
112 clear_buffer_pinned(bh); 80 clear_buffer_pinned(bh);
113 81
114 if (buffer_is_rgrp(bd)) !! 82 gfs2_log_lock(sdp);
115 maybe_release_space(bd); !! 83 if (bd->bd_ail) {
116 <<
117 spin_lock(&sdp->sd_ail_lock); <<
118 if (bd->bd_tr) { <<
119 list_del(&bd->bd_ail_st_list); 84 list_del(&bd->bd_ail_st_list);
120 brelse(bh); 85 brelse(bh);
121 } else { 86 } else {
122 struct gfs2_glock *gl = bd->bd 87 struct gfs2_glock *gl = bd->bd_gl;
123 list_add(&bd->bd_ail_gl_list, 88 list_add(&bd->bd_ail_gl_list, &gl->gl_ail_list);
124 atomic_inc(&gl->gl_ail_count); 89 atomic_inc(&gl->gl_ail_count);
125 } 90 }
126 bd->bd_tr = tr; !! 91 bd->bd_ail = ai;
127 list_add(&bd->bd_ail_st_list, &tr->tr_ !! 92 list_add(&bd->bd_ail_st_list, &ai->ai_ail1_list);
128 spin_unlock(&sdp->sd_ail_lock); <<
129 <<
130 clear_bit(GLF_LFLUSH, &bd->bd_gl->gl_f 93 clear_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags);
131 trace_gfs2_pin(bd, 0); 94 trace_gfs2_pin(bd, 0);
>> 95 gfs2_log_unlock(sdp);
132 unlock_buffer(bh); 96 unlock_buffer(bh);
133 atomic_dec(&sdp->sd_log_pinned); <<
134 } 97 }
135 98
136 void gfs2_log_incr_head(struct gfs2_sbd *sdp) <<
137 { <<
138 BUG_ON((sdp->sd_log_flush_head == sdp- <<
139 (sdp->sd_log_flush_head != sdp- <<
140 99
141 if (++sdp->sd_log_flush_head == sdp->s !! 100 static inline struct gfs2_log_descriptor *bh_log_desc(struct buffer_head *bh)
142 sdp->sd_log_flush_head = 0; <<
143 } <<
144 <<
145 u64 gfs2_log_bmap(struct gfs2_jdesc *jd, unsig <<
146 { 101 {
147 struct gfs2_journal_extent *je; !! 102 return (struct gfs2_log_descriptor *)bh->b_data;
148 <<
149 list_for_each_entry(je, &jd->extent_li <<
150 if (lblock >= je->lblock && lb <<
151 return je->dblock + lb <<
152 } <<
153 <<
154 return -1; <<
155 } 103 }
156 104
157 /** !! 105 static inline __be64 *bh_log_ptr(struct buffer_head *bh)
158 * gfs2_end_log_write_bh - end log write of pa <<
159 * @sdp: The superblock <<
160 * @bvec: The bio_vec <<
161 * @error: The i/o status <<
162 * <<
163 * This finds the relevant buffers and unlocks <<
164 * error flag according to the status of the i <<
165 * used when the log is writing data which has <<
166 * that is pinned in the pagecache. <<
167 */ <<
168 <<
169 static void gfs2_end_log_write_bh(struct gfs2_ <<
170 struct bio_v <<
171 blk_status_t <<
172 { <<
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) <<
183 mark_buffer_write_io_e <<
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); <<
190 } <<
191 <<
192 /** <<
193 * gfs2_end_log_write - end of i/o to the log <<
194 * @bio: The bio <<
195 * <<
196 * Each bio_vec contains either data from the <<
197 * relating to the log itself. Here we iterate <<
198 * array, processing both kinds of data. <<
199 * <<
200 */ <<
201 <<
202 static void gfs2_end_log_write(struct bio *bio <<
203 { 106 {
204 struct gfs2_sbd *sdp = bio->bi_private !! 107 struct gfs2_log_descriptor *ld = bh_log_desc(bh);
205 struct bio_vec *bvec; !! 108 return (__force __be64 *)(ld + 1);
206 struct page *page; <<
207 struct bvec_iter_all iter_all; <<
208 <<
209 if (bio->bi_status) { <<
210 if (!cmpxchg(&sdp->sd_log_erro <<
211 fs_err(sdp, "Error %d <<
212 bio->bi_status, <<
213 gfs2_withdraw_delayed(sdp); <<
214 /* prevent more writes to the <<
215 clear_bit(SDF_JOURNAL_LIVE, &s <<
216 wake_up(&sdp->sd_logd_waitq); <<
217 } <<
218 <<
219 bio_for_each_segment_all(bvec, bio, it <<
220 page = bvec->bv_page; <<
221 if (page_has_buffers(page)) <<
222 gfs2_end_log_write_bh( <<
223 else <<
224 mempool_free(page, gfs <<
225 } <<
226 <<
227 bio_put(bio); <<
228 if (atomic_dec_and_test(&sdp->sd_log_i <<
229 wake_up(&sdp->sd_log_flush_wai <<
230 } 109 }
231 110
232 /** !! 111 static inline __be64 *bh_ptr_end(struct buffer_head *bh)
233 * gfs2_log_submit_bio - Submit any pending lo <<
234 * @biop: Address of the bio pointer <<
235 * @opf: REQ_OP | op_flags <<
236 * <<
237 * Submit any pending part-built or full bio t <<
238 * there is no pending bio, then this is a no- <<
239 */ <<
240 <<
241 void gfs2_log_submit_bio(struct bio **biop, bl <<
242 { 112 {
243 struct bio *bio = *biop; !! 113 return (__force __be64 *)(bh->b_data + bh->b_size);
244 if (bio) { <<
245 struct gfs2_sbd *sdp = bio->bi <<
246 atomic_inc(&sdp->sd_log_in_fli <<
247 bio->bi_opf = opf; <<
248 submit_bio(bio); <<
249 *biop = NULL; <<
250 } <<
251 } <<
252 <<
253 /** <<
254 * gfs2_log_alloc_bio - Allocate a bio <<
255 * @sdp: The super block <<
256 * @blkno: The device block number we want to <<
257 * @end_io: The bi_end_io callback <<
258 * <<
259 * Allocate a new bio, initialize it with the <<
260 * <<
261 * Returns: The newly allocated bio <<
262 */ <<
263 <<
264 static struct bio *gfs2_log_alloc_bio(struct g <<
265 bio_end_ <<
266 { <<
267 struct super_block *sb = sdp->sd_vfs; <<
268 struct bio *bio = bio_alloc(sb->s_bdev <<
269 <<
270 bio->bi_iter.bi_sector = blkno << sdp- <<
271 bio->bi_end_io = end_io; <<
272 bio->bi_private = sdp; <<
273 <<
274 return bio; <<
275 } 114 }
276 115
277 /** <<
278 * gfs2_log_get_bio - Get cached log bio, or a <<
279 * @sdp: The super block <<
280 * @blkno: The device block number we want to <<
281 * @biop: The bio to get or allocate <<
282 * @op: REQ_OP <<
283 * @end_io: The bi_end_io callback <<
284 * @flush: Always flush the current bio and al <<
285 * <<
286 * If there is a cached bio, then if the next <<
287 * with the previous one, return it, otherwise <<
288 * device. If there is no cached bio, or we ju <<
289 * allocate a new one. <<
290 * <<
291 * Returns: The bio to use for log writes <<
292 */ <<
293 <<
294 static struct bio *gfs2_log_get_bio(struct gfs <<
295 struct bio <<
296 bio_end_io <<
297 { <<
298 struct bio *bio = *biop; <<
299 <<
300 if (bio) { <<
301 u64 nblk; <<
302 <<
303 nblk = bio_end_sector(bio); <<
304 nblk >>= sdp->sd_fsb2bb_shift; <<
305 if (blkno == nblk && !flush) <<
306 return bio; <<
307 gfs2_log_submit_bio(biop, op); <<
308 } <<
309 <<
310 *biop = gfs2_log_alloc_bio(sdp, blkno, <<
311 return *biop; <<
312 } <<
313 <<
314 /** <<
315 * gfs2_log_write - write to log <<
316 * @sdp: the filesystem <<
317 * @jd: The journal descriptor <<
318 * @page: the page to write <<
319 * @size: the size of the data to write <<
320 * @offset: the offset within the page <<
321 * @blkno: block number of the log entry <<
322 * <<
323 * Try and add the page segment to the current <<
324 * submit the current bio to the device and cr <<
325 * then add the page segment to that. <<
326 */ <<
327 <<
328 void gfs2_log_write(struct gfs2_sbd *sdp, stru <<
329 struct page *page, unsigne <<
330 u64 blkno) <<
331 { <<
332 struct bio *bio; <<
333 int ret; <<
334 <<
335 bio = gfs2_log_get_bio(sdp, blkno, &jd <<
336 gfs2_end_log_wr <<
337 ret = bio_add_page(bio, page, size, of <<
338 if (ret == 0) { <<
339 bio = gfs2_log_get_bio(sdp, bl <<
340 REQ_OP_ <<
341 ret = bio_add_page(bio, page, <<
342 WARN_ON(ret == 0); <<
343 } <<
344 } <<
345 <<
346 /** <<
347 * gfs2_log_write_bh - write a buffer's conten <<
348 * @sdp: The super block <<
349 * @bh: The buffer pointing to the in-place lo <<
350 * <<
351 * This writes the content of the buffer to th <<
352 * in the log. The buffer will be unlocked onc <<
353 * completed. <<
354 */ <<
355 <<
356 static void gfs2_log_write_bh(struct gfs2_sbd <<
357 { <<
358 u64 dblock; <<
359 <<
360 dblock = gfs2_log_bmap(sdp->sd_jdesc, <<
361 gfs2_log_incr_head(sdp); <<
362 gfs2_log_write(sdp, sdp->sd_jdesc, bh- <<
363 bh_offset(bh), dblock); <<
364 } <<
365 <<
366 /** <<
367 * gfs2_log_write_page - write one block store <<
368 * @sdp: The superblock <<
369 * @page: The struct page <<
370 * <<
371 * This writes the first block-sized part of t <<
372 * that the page must have been allocated from <<
373 * and that after this has been called, owners <<
374 * the page may be freed at any time. <<
375 */ <<
376 <<
377 static void gfs2_log_write_page(struct gfs2_sb <<
378 { <<
379 struct super_block *sb = sdp->sd_vfs; <<
380 u64 dblock; <<
381 <<
382 dblock = gfs2_log_bmap(sdp->sd_jdesc, <<
383 gfs2_log_incr_head(sdp); <<
384 gfs2_log_write(sdp, sdp->sd_jdesc, pag <<
385 } <<
386 <<
387 /** <<
388 * gfs2_end_log_read - end I/O callback for re <<
389 * @bio: The bio <<
390 * <<
391 * Simply unlock the pages in the bio. The mai <<
392 * process them in order as necessary. <<
393 */ <<
394 static void gfs2_end_log_read(struct bio *bio) <<
395 { <<
396 int error = blk_status_to_errno(bio->b <<
397 struct folio_iter fi; <<
398 <<
399 bio_for_each_folio_all(fi, bio) { <<
400 /* We're abusing wb_err to get <<
401 filemap_set_wb_err(fi.folio->m <<
402 folio_end_read(fi.folio, !erro <<
403 } <<
404 <<
405 bio_put(bio); <<
406 } <<
407 <<
408 /** <<
409 * gfs2_jhead_pg_srch - Look for the journal h <<
410 * @jd: The journal descriptor <<
411 * @head: The journal head to start from <<
412 * @page: The page to look in <<
413 * <<
414 * Returns: 1 if found, 0 otherwise. <<
415 */ <<
416 <<
417 static bool gfs2_jhead_pg_srch(struct gfs2_jde <<
418 struct gfs2_log_ <<
419 struct page *pag <<
420 { <<
421 struct gfs2_sbd *sdp = GFS2_SB(jd->jd_ <<
422 struct gfs2_log_header_host lh; <<
423 void *kaddr; <<
424 unsigned int offset; <<
425 bool ret = false; <<
426 <<
427 kaddr = kmap_local_page(page); <<
428 for (offset = 0; offset < PAGE_SIZE; o <<
429 if (!__get_log_header(sdp, kad <<
430 if (lh.lh_sequence >= <<
431 *head = lh; <<
432 else { <<
433 ret = true; <<
434 break; <<
435 } <<
436 } <<
437 } <<
438 kunmap_local(kaddr); <<
439 return ret; <<
440 } <<
441 <<
442 /** <<
443 * gfs2_jhead_process_page - Search/cleanup a <<
444 * @jd: The journal descriptor <<
445 * @index: Index of the page to look into <<
446 * @head: The journal head to start from <<
447 * @done: If set, perform only cleanup, else s <<
448 * <<
449 * Find the folio with 'index' in the journal' <<
450 * the journal head if requested (cleanup == f <<
451 * folio so the page cache can reclaim it. We <<
452 * reference on this folio twice, first when w <<
453 * to obtain the folio to add it to the bio an <<
454 * filemap_get_folio() here to get the folio t <<
455 * completed. <<
456 * This function is also used to free up a fol <<
457 * used. Maybe we added it to a bio, but not s <<
458 * submitted the I/O, but we already found the <<
459 * our references to the folio. <<
460 */ <<
461 <<
462 static void gfs2_jhead_process_page(struct gfs <<
463 struct gfs <<
464 bool *done <<
465 { <<
466 struct folio *folio; <<
467 <<
468 folio = filemap_get_folio(jd->jd_inode <<
469 <<
470 folio_wait_locked(folio); <<
471 if (!folio_test_uptodate(folio)) <<
472 *done = true; <<
473 <<
474 if (!*done) <<
475 *done = gfs2_jhead_pg_srch(jd, <<
476 <<
477 /* filemap_get_folio() and the earlier <<
478 folio_put_refs(folio, 2); <<
479 } <<
480 <<
481 static struct bio *gfs2_chain_bio(struct bio * <<
482 { <<
483 struct bio *new; <<
484 <<
485 new = bio_alloc(prev->bi_bdev, nr_iove <<
486 bio_clone_blkg_association(new, prev); <<
487 new->bi_iter.bi_sector = bio_end_secto <<
488 bio_chain(new, prev); <<
489 submit_bio(prev); <<
490 return new; <<
491 } <<
492 <<
493 /** <<
494 * gfs2_find_jhead - find the head of a log <<
495 * @jd: The journal descriptor <<
496 * @head: The log descriptor for the head of t <<
497 * @keep_cache: If set inode pages will not be <<
498 * <<
499 * Do a search of a journal by reading it in l <<
500 * the valid log entry with the highest sequen <<
501 * <<
502 * Returns: 0 on success, errno otherwise <<
503 */ <<
504 int gfs2_find_jhead(struct gfs2_jdesc *jd, str <<
505 bool keep_cache) <<
506 { <<
507 struct gfs2_sbd *sdp = GFS2_SB(jd->jd_ <<
508 struct address_space *mapping = jd->jd <<
509 unsigned int block = 0, blocks_submitt <<
510 unsigned int bsize = sdp->sd_sb.sb_bsi <<
511 unsigned int bsize_shift = sdp->sd_sb. <<
512 unsigned int shift = PAGE_SHIFT - bsiz <<
513 unsigned int max_blocks = 2 * 1024 * 1 <<
514 struct gfs2_journal_extent *je; <<
515 int sz, ret = 0; <<
516 struct bio *bio = NULL; <<
517 struct page *page = NULL; <<
518 bool done = false; <<
519 errseq_t since; <<
520 <<
521 memset(head, 0, sizeof(*head)); <<
522 if (list_empty(&jd->extent_list)) <<
523 gfs2_map_journal_extents(sdp, <<
524 <<
525 since = filemap_sample_wb_err(mapping) <<
526 list_for_each_entry(je, &jd->extent_li <<
527 u64 dblock = je->dblock; <<
528 <<
529 for (; block < je->lblock + je <<
530 if (!page) { <<
531 page = grab_ca <<
532 if (!page) { <<
533 ret = <<
534 done = <<
535 goto o <<
536 } <<
537 off = 0; <<
538 } <<
539 <<
540 if (bio && (off || blo <<
541 sector_t secto <<
542 <<
543 if (bio_end_se <<
544 sz = b <<
545 if (sz <<
546 <<
547 } <<
548 if (off) { <<
549 unsign <<
550 <<
551 <<
552 bio = <<
553 goto a <<
554 } <<
555 } <<
556 <<
557 if (bio) { <<
558 blocks_submitt <<
559 submit_bio(bio <<
560 } <<
561 <<
562 bio = gfs2_log_alloc_b <<
563 bio->bi_opf = REQ_OP_R <<
564 add_block_to_new_bio: <<
565 sz = bio_add_page(bio, <<
566 BUG_ON(sz != bsize); <<
567 block_added: <<
568 off += bsize; <<
569 if (off == PAGE_SIZE) <<
570 page = NULL; <<
571 if (blocks_submitted < <<
572 /* Keep at lea <<
573 continue; <<
574 } <<
575 <<
576 gfs2_jhead_process_pag <<
577 blocks_read += PAGE_SI <<
578 if (done) <<
579 goto out; /* <<
580 } <<
581 } <<
582 <<
583 out: <<
584 if (bio) <<
585 submit_bio(bio); <<
586 while (blocks_read < block) { <<
587 gfs2_jhead_process_page(jd, bl <<
588 blocks_read += PAGE_SIZE >> bs <<
589 } <<
590 <<
591 if (!ret) <<
592 ret = filemap_check_wb_err(map <<
593 <<
594 if (!keep_cache) <<
595 truncate_inode_pages(mapping, <<
596 <<
597 return ret; <<
598 } <<
599 116
600 static struct page *gfs2_get_log_desc(struct g !! 117 static struct buffer_head *gfs2_get_log_desc(struct gfs2_sbd *sdp, u32 ld_type)
601 u32 ld_l <<
602 { 118 {
603 struct page *page = mempool_alloc(gfs2 !! 119 struct buffer_head *bh = gfs2_log_get_buf(sdp);
604 struct gfs2_log_descriptor *ld = page_ !! 120 struct gfs2_log_descriptor *ld = bh_log_desc(bh);
605 clear_page(ld); <<
606 ld->ld_header.mh_magic = cpu_to_be32(G 121 ld->ld_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
607 ld->ld_header.mh_type = cpu_to_be32(GF 122 ld->ld_header.mh_type = cpu_to_be32(GFS2_METATYPE_LD);
608 ld->ld_header.mh_format = cpu_to_be32( 123 ld->ld_header.mh_format = cpu_to_be32(GFS2_FORMAT_LD);
609 ld->ld_type = cpu_to_be32(ld_type); 124 ld->ld_type = cpu_to_be32(ld_type);
610 ld->ld_length = cpu_to_be32(ld_length) !! 125 ld->ld_length = 0;
611 ld->ld_data1 = cpu_to_be32(ld_data1); !! 126 ld->ld_data1 = 0;
612 ld->ld_data2 = 0; 127 ld->ld_data2 = 0;
613 return page; !! 128 memset(ld->ld_reserved, 0, sizeof(ld->ld_reserved));
>> 129 return bh;
614 } 130 }
615 131
616 static void gfs2_check_magic(struct buffer_hea !! 132 static void buf_lo_add(struct gfs2_sbd *sdp, struct gfs2_log_element *le)
617 { <<
618 void *kaddr; <<
619 __be32 *ptr; <<
620 <<
621 clear_buffer_escaped(bh); <<
622 kaddr = kmap_local_page(bh->b_page); <<
623 ptr = kaddr + bh_offset(bh); <<
624 if (*ptr == cpu_to_be32(GFS2_MAGIC)) <<
625 set_buffer_escaped(bh); <<
626 kunmap_local(kaddr); <<
627 } <<
628 <<
629 static int blocknr_cmp(void *priv, const struc <<
630 const struct list_head <<
631 { 133 {
632 struct gfs2_bufdata *bda, *bdb; !! 134 struct gfs2_bufdata *bd = container_of(le, struct gfs2_bufdata, bd_le);
633 !! 135 struct gfs2_trans *tr;
634 bda = list_entry(a, struct gfs2_bufdat <<
635 bdb = list_entry(b, struct gfs2_bufdat <<
636 136
637 if (bda->bd_bh->b_blocknr < bdb->bd_bh !! 137 lock_buffer(bd->bd_bh);
638 return -1; !! 138 gfs2_log_lock(sdp);
639 if (bda->bd_bh->b_blocknr > bdb->bd_bh !! 139 if (!list_empty(&bd->bd_list_tr))
640 return 1; !! 140 goto out;
641 return 0; !! 141 tr = current->journal_info;
>> 142 tr->tr_touched = 1;
>> 143 tr->tr_num_buf++;
>> 144 list_add(&bd->bd_list_tr, &tr->tr_list_buf);
>> 145 if (!list_empty(&le->le_list))
>> 146 goto out;
>> 147 set_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags);
>> 148 set_bit(GLF_DIRTY, &bd->bd_gl->gl_flags);
>> 149 gfs2_meta_check(sdp, bd->bd_bh);
>> 150 gfs2_pin(sdp, bd->bd_bh);
>> 151 sdp->sd_log_num_buf++;
>> 152 list_add(&le->le_list, &sdp->sd_log_le_buf);
>> 153 tr->tr_num_buf_new++;
>> 154 out:
>> 155 gfs2_log_unlock(sdp);
>> 156 unlock_buffer(bd->bd_bh);
642 } 157 }
643 158
644 static void gfs2_before_commit(struct gfs2_sbd !! 159 static void buf_lo_before_commit(struct gfs2_sbd *sdp)
645 unsigned int t <<
646 bool is_databu <<
647 { 160 {
>> 161 struct buffer_head *bh;
648 struct gfs2_log_descriptor *ld; 162 struct gfs2_log_descriptor *ld;
649 struct gfs2_bufdata *bd1 = NULL, *bd2; 163 struct gfs2_bufdata *bd1 = NULL, *bd2;
650 struct page *page; !! 164 unsigned int total;
>> 165 unsigned int limit;
651 unsigned int num; 166 unsigned int num;
652 unsigned n; 167 unsigned n;
653 __be64 *ptr; 168 __be64 *ptr;
654 169
>> 170 limit = buf_limit(sdp);
>> 171 /* for 4k blocks, limit = 503 */
>> 172
655 gfs2_log_lock(sdp); 173 gfs2_log_lock(sdp);
656 list_sort(NULL, blist, blocknr_cmp); !! 174 total = sdp->sd_log_num_buf;
657 bd1 = bd2 = list_prepare_entry(bd1, bl !! 175 bd1 = bd2 = list_prepare_entry(bd1, &sdp->sd_log_le_buf, bd_le.le_list);
658 while(total) { 176 while(total) {
659 num = total; 177 num = total;
660 if (total > limit) 178 if (total > limit)
661 num = limit; 179 num = limit;
662 gfs2_log_unlock(sdp); 180 gfs2_log_unlock(sdp);
663 page = gfs2_get_log_desc(sdp, !! 181 bh = gfs2_get_log_desc(sdp, GFS2_LOG_DESC_METADATA);
664 is_da <<
665 GFS2_ <<
666 ld = page_address(page); <<
667 gfs2_log_lock(sdp); 182 gfs2_log_lock(sdp);
668 ptr = (__be64 *)(ld + 1); !! 183 ld = bh_log_desc(bh);
>> 184 ptr = bh_log_ptr(bh);
>> 185 ld->ld_length = cpu_to_be32(num + 1);
>> 186 ld->ld_data1 = cpu_to_be32(num);
669 187
670 n = 0; 188 n = 0;
671 list_for_each_entry_continue(b !! 189 list_for_each_entry_continue(bd1, &sdp->sd_log_le_buf,
>> 190 bd_le.le_list) {
672 *ptr++ = cpu_to_be64(b 191 *ptr++ = cpu_to_be64(bd1->bd_bh->b_blocknr);
673 if (is_databuf) { <<
674 gfs2_check_mag <<
675 *ptr++ = cpu_t <<
676 } <<
677 if (++n >= num) 192 if (++n >= num)
678 break; 193 break;
679 } 194 }
680 195
681 gfs2_log_unlock(sdp); 196 gfs2_log_unlock(sdp);
682 gfs2_log_write_page(sdp, page) !! 197 submit_bh(WRITE_SYNC_PLUG, bh);
683 gfs2_log_lock(sdp); 198 gfs2_log_lock(sdp);
684 199
685 n = 0; 200 n = 0;
686 list_for_each_entry_continue(b !! 201 list_for_each_entry_continue(bd2, &sdp->sd_log_le_buf,
>> 202 bd_le.le_list) {
687 get_bh(bd2->bd_bh); 203 get_bh(bd2->bd_bh);
688 gfs2_log_unlock(sdp); 204 gfs2_log_unlock(sdp);
689 lock_buffer(bd2->bd_bh 205 lock_buffer(bd2->bd_bh);
690 !! 206 bh = gfs2_log_fake_buf(sdp, bd2->bd_bh);
691 if (buffer_escaped(bd2 !! 207 submit_bh(WRITE_SYNC_PLUG, bh);
692 void *p; <<
693 <<
694 page = mempool <<
695 p = page_addre <<
696 memcpy_from_pa <<
697 *(__be32 *)p = <<
698 clear_buffer_e <<
699 unlock_buffer( <<
700 brelse(bd2->bd <<
701 gfs2_log_write <<
702 } else { <<
703 gfs2_log_write <<
704 } <<
705 gfs2_log_lock(sdp); 208 gfs2_log_lock(sdp);
706 if (++n >= num) 209 if (++n >= num)
707 break; 210 break;
708 } 211 }
709 212
710 BUG_ON(total < num); 213 BUG_ON(total < num);
711 total -= num; 214 total -= num;
712 } 215 }
713 gfs2_log_unlock(sdp); 216 gfs2_log_unlock(sdp);
714 } 217 }
715 218
716 static void buf_lo_before_commit(struct gfs2_s !! 219 static void buf_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_ail *ai)
717 { <<
718 unsigned int limit = buf_limit(sdp); / <<
719 unsigned int nbuf; <<
720 if (tr == NULL) <<
721 return; <<
722 nbuf = tr->tr_num_buf_new - tr->tr_num <<
723 gfs2_before_commit(sdp, limit, nbuf, & <<
724 } <<
725 <<
726 static void buf_lo_after_commit(struct gfs2_sb <<
727 { 220 {
728 struct list_head *head; !! 221 struct list_head *head = &sdp->sd_log_le_buf;
729 struct gfs2_bufdata *bd; 222 struct gfs2_bufdata *bd;
730 223
731 if (tr == NULL) <<
732 return; <<
733 <<
734 head = &tr->tr_buf; <<
735 while (!list_empty(head)) { 224 while (!list_empty(head)) {
736 bd = list_first_entry(head, st !! 225 bd = list_entry(head->next, struct gfs2_bufdata, bd_le.le_list);
737 list_del_init(&bd->bd_list); !! 226 list_del_init(&bd->bd_le.le_list);
738 gfs2_unpin(sdp, bd->bd_bh, tr) !! 227 sdp->sd_log_num_buf--;
>> 228
>> 229 gfs2_unpin(sdp, bd->bd_bh, ai);
739 } 230 }
>> 231 gfs2_assert_warn(sdp, !sdp->sd_log_num_buf);
740 } 232 }
741 233
742 static void buf_lo_before_scan(struct gfs2_jde 234 static void buf_lo_before_scan(struct gfs2_jdesc *jd,
743 struct gfs2_log 235 struct gfs2_log_header_host *head, int pass)
744 { 236 {
>> 237 struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
>> 238
745 if (pass != 0) 239 if (pass != 0)
746 return; 240 return;
747 241
748 jd->jd_found_blocks = 0; !! 242 sdp->sd_found_blocks = 0;
749 jd->jd_replayed_blocks = 0; !! 243 sdp->sd_replayed_blocks = 0;
750 } 244 }
751 245
752 #define obsolete_rgrp_replay \ !! 246 static int buf_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start,
753 "Replaying 0x%llx from jid=%d/0x%llx but we al <<
754 #define obsolete_rgrp_replay2 \ <<
755 "busy:%d, pinned:%d rg_gen:0x%llx, j_gen:0x%ll <<
756 <<
757 static void obsolete_rgrp(struct gfs2_jdesc *j <<
758 u64 blkno) <<
759 { <<
760 struct gfs2_sbd *sdp = GFS2_SB(jd->jd_ <<
761 struct gfs2_rgrpd *rgd; <<
762 struct gfs2_rgrp *jrgd = (struct gfs2_ <<
763 <<
764 rgd = gfs2_blk2rgrpd(sdp, blkno, false <<
765 if (rgd && rgd->rd_addr == blkno && <<
766 rgd->rd_bits && rgd->rd_bits->bi_b <<
767 fs_info(sdp, obsolete_rgrp_rep <<
768 jd->jd_jid, bh_log->b_ <<
769 fs_info(sdp, obsolete_rgrp_rep <<
770 buffer_busy(rgd->rd_bi <<
771 buffer_pinned(rgd->rd_ <<
772 rgd->rd_igeneration, <<
773 be64_to_cpu(jrgd->rg_i <<
774 gfs2_dump_glock(NULL, rgd->rd_ <<
775 } <<
776 } <<
777 <<
778 static int buf_lo_scan_elements(struct gfs2_jd <<
779 struct gfs2_lo 247 struct gfs2_log_descriptor *ld, __be64 *ptr,
780 int pass) 248 int pass)
781 { 249 {
782 struct gfs2_inode *ip = GFS2_I(jd->jd_ 250 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
783 struct gfs2_sbd *sdp = GFS2_SB(jd->jd_ 251 struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
784 struct gfs2_glock *gl = ip->i_gl; 252 struct gfs2_glock *gl = ip->i_gl;
785 unsigned int blks = be32_to_cpu(ld->ld 253 unsigned int blks = be32_to_cpu(ld->ld_data1);
786 struct buffer_head *bh_log, *bh_ip; 254 struct buffer_head *bh_log, *bh_ip;
787 u64 blkno; 255 u64 blkno;
788 int error = 0; 256 int error = 0;
789 257
790 if (pass != 1 || be32_to_cpu(ld->ld_ty 258 if (pass != 1 || be32_to_cpu(ld->ld_type) != GFS2_LOG_DESC_METADATA)
791 return 0; 259 return 0;
792 260
793 gfs2_replay_incr_blk(jd, &start); !! 261 gfs2_replay_incr_blk(sdp, &start);
794 262
795 for (; blks; gfs2_replay_incr_blk(jd, !! 263 for (; blks; gfs2_replay_incr_blk(sdp, &start), blks--) {
796 blkno = be64_to_cpu(*ptr++); 264 blkno = be64_to_cpu(*ptr++);
797 265
798 jd->jd_found_blocks++; !! 266 sdp->sd_found_blocks++;
799 267
800 if (gfs2_revoke_check(jd, blkn !! 268 if (gfs2_revoke_check(sdp, blkno, start))
801 continue; 269 continue;
802 270
803 error = gfs2_replay_read_block 271 error = gfs2_replay_read_block(jd, start, &bh_log);
804 if (error) 272 if (error)
805 return error; 273 return error;
806 274
807 bh_ip = gfs2_meta_new(gl, blkn 275 bh_ip = gfs2_meta_new(gl, blkno);
808 memcpy(bh_ip->b_data, bh_log-> 276 memcpy(bh_ip->b_data, bh_log->b_data, bh_log->b_size);
809 277
810 if (gfs2_meta_check(sdp, bh_ip 278 if (gfs2_meta_check(sdp, bh_ip))
811 error = -EIO; 279 error = -EIO;
812 else { !! 280 else
813 struct gfs2_meta_heade <<
814 (struct gfs2_m <<
815 <<
816 if (mh->mh_type == cpu <<
817 obsolete_rgrp( <<
818 <<
819 mark_buffer_dirty(bh_i 281 mark_buffer_dirty(bh_ip);
820 } !! 282
821 brelse(bh_log); 283 brelse(bh_log);
822 brelse(bh_ip); 284 brelse(bh_ip);
823 285
824 if (error) 286 if (error)
825 break; 287 break;
826 288
827 jd->jd_replayed_blocks++; !! 289 sdp->sd_replayed_blocks++;
828 } 290 }
829 291
830 return error; 292 return error;
831 } 293 }
832 294
833 static void buf_lo_after_scan(struct gfs2_jdes 295 static void buf_lo_after_scan(struct gfs2_jdesc *jd, int error, int pass)
834 { 296 {
835 struct gfs2_inode *ip = GFS2_I(jd->jd_ 297 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
836 struct gfs2_sbd *sdp = GFS2_SB(jd->jd_ 298 struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
837 299
838 if (error) { 300 if (error) {
839 gfs2_inode_metasync(ip->i_gl); !! 301 gfs2_meta_sync(ip->i_gl);
840 return; 302 return;
841 } 303 }
842 if (pass != 1) 304 if (pass != 1)
843 return; 305 return;
844 306
845 gfs2_inode_metasync(ip->i_gl); !! 307 gfs2_meta_sync(ip->i_gl);
846 308
847 fs_info(sdp, "jid=%u: Replayed %u of % 309 fs_info(sdp, "jid=%u: Replayed %u of %u blocks\n",
848 jd->jd_jid, jd->jd_replayed_bl !! 310 jd->jd_jid, sdp->sd_replayed_blocks, sdp->sd_found_blocks);
849 } 311 }
850 312
851 static void revoke_lo_before_commit(struct gfs !! 313 static void revoke_lo_add(struct gfs2_sbd *sdp, struct gfs2_log_element *le)
852 { 314 {
>> 315 struct gfs2_trans *tr;
>> 316
>> 317 tr = current->journal_info;
>> 318 tr->tr_touched = 1;
>> 319 tr->tr_num_revoke++;
>> 320 sdp->sd_log_num_revoke++;
>> 321 list_add(&le->le_list, &sdp->sd_log_le_revoke);
>> 322 }
>> 323
>> 324 static void revoke_lo_before_commit(struct gfs2_sbd *sdp)
>> 325 {
>> 326 struct gfs2_log_descriptor *ld;
853 struct gfs2_meta_header *mh; 327 struct gfs2_meta_header *mh;
>> 328 struct buffer_head *bh;
854 unsigned int offset; 329 unsigned int offset;
855 struct list_head *head = &sdp->sd_log_ !! 330 struct list_head *head = &sdp->sd_log_le_revoke;
856 struct gfs2_bufdata *bd; 331 struct gfs2_bufdata *bd;
857 struct page *page; <<
858 unsigned int length; <<
859 332
860 gfs2_flush_revokes(sdp); <<
861 if (!sdp->sd_log_num_revoke) 333 if (!sdp->sd_log_num_revoke)
862 return; 334 return;
863 335
864 length = gfs2_struct2blk(sdp, sdp->sd_ !! 336 bh = gfs2_get_log_desc(sdp, GFS2_LOG_DESC_REVOKE);
865 page = gfs2_get_log_desc(sdp, GFS2_LOG !! 337 ld = bh_log_desc(bh);
>> 338 ld->ld_length = cpu_to_be32(gfs2_struct2blk(sdp, sdp->sd_log_num_revoke,
>> 339 sizeof(u64)));
>> 340 ld->ld_data1 = cpu_to_be32(sdp->sd_log_num_revoke);
866 offset = sizeof(struct gfs2_log_descri 341 offset = sizeof(struct gfs2_log_descriptor);
867 342
868 list_for_each_entry(bd, head, bd_list) !! 343 while (!list_empty(head)) {
>> 344 bd = list_entry(head->next, struct gfs2_bufdata, bd_le.le_list);
>> 345 list_del_init(&bd->bd_le.le_list);
869 sdp->sd_log_num_revoke--; 346 sdp->sd_log_num_revoke--;
870 347
871 if (offset + sizeof(u64) > sdp 348 if (offset + sizeof(u64) > sdp->sd_sb.sb_bsize) {
872 gfs2_log_write_page(sd !! 349 submit_bh(WRITE_SYNC_PLUG, bh);
873 page = mempool_alloc(g !! 350
874 mh = page_address(page !! 351 bh = gfs2_log_get_buf(sdp);
875 clear_page(mh); !! 352 mh = (struct gfs2_meta_header *)bh->b_data;
876 mh->mh_magic = cpu_to_ 353 mh->mh_magic = cpu_to_be32(GFS2_MAGIC);
877 mh->mh_type = cpu_to_b 354 mh->mh_type = cpu_to_be32(GFS2_METATYPE_LB);
878 mh->mh_format = cpu_to 355 mh->mh_format = cpu_to_be32(GFS2_FORMAT_LB);
879 offset = sizeof(struct 356 offset = sizeof(struct gfs2_meta_header);
880 } 357 }
881 358
882 *(__be64 *)(page_address(page) !! 359 *(__be64 *)(bh->b_data + offset) = cpu_to_be64(bd->bd_blkno);
>> 360 kmem_cache_free(gfs2_bufdata_cachep, bd);
>> 361
883 offset += sizeof(u64); 362 offset += sizeof(u64);
884 } 363 }
885 gfs2_assert_withdraw(sdp, !sdp->sd_log 364 gfs2_assert_withdraw(sdp, !sdp->sd_log_num_revoke);
886 365
887 gfs2_log_write_page(sdp, page); !! 366 submit_bh(WRITE_SYNC_PLUG, bh);
888 } <<
889 <<
890 void gfs2_drain_revokes(struct gfs2_sbd *sdp) <<
891 { <<
892 struct list_head *head = &sdp->sd_log_ <<
893 struct gfs2_bufdata *bd; <<
894 struct gfs2_glock *gl; <<
895 <<
896 while (!list_empty(head)) { <<
897 bd = list_first_entry(head, st <<
898 list_del_init(&bd->bd_list); <<
899 gl = bd->bd_gl; <<
900 gfs2_glock_remove_revoke(gl); <<
901 kmem_cache_free(gfs2_bufdata_c <<
902 } <<
903 } <<
904 <<
905 static void revoke_lo_after_commit(struct gfs2 <<
906 { <<
907 gfs2_drain_revokes(sdp); <<
908 } 367 }
909 368
910 static void revoke_lo_before_scan(struct gfs2_ 369 static void revoke_lo_before_scan(struct gfs2_jdesc *jd,
911 struct gfs2_ 370 struct gfs2_log_header_host *head, int pass)
912 { 371 {
>> 372 struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
>> 373
913 if (pass != 0) 374 if (pass != 0)
914 return; 375 return;
915 376
916 jd->jd_found_revokes = 0; !! 377 sdp->sd_found_revokes = 0;
917 jd->jd_replay_tail = head->lh_tail; !! 378 sdp->sd_replay_tail = head->lh_tail;
918 } 379 }
919 380
920 static int revoke_lo_scan_elements(struct gfs2 !! 381 static int revoke_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start,
921 struct gfs2 382 struct gfs2_log_descriptor *ld, __be64 *ptr,
922 int pass) 383 int pass)
923 { 384 {
924 struct gfs2_sbd *sdp = GFS2_SB(jd->jd_ 385 struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
925 unsigned int blks = be32_to_cpu(ld->ld 386 unsigned int blks = be32_to_cpu(ld->ld_length);
926 unsigned int revokes = be32_to_cpu(ld- 387 unsigned int revokes = be32_to_cpu(ld->ld_data1);
927 struct buffer_head *bh; 388 struct buffer_head *bh;
928 unsigned int offset; 389 unsigned int offset;
929 u64 blkno; 390 u64 blkno;
930 int first = 1; 391 int first = 1;
931 int error; 392 int error;
932 393
933 if (pass != 0 || be32_to_cpu(ld->ld_ty 394 if (pass != 0 || be32_to_cpu(ld->ld_type) != GFS2_LOG_DESC_REVOKE)
934 return 0; 395 return 0;
935 396
936 offset = sizeof(struct gfs2_log_descri 397 offset = sizeof(struct gfs2_log_descriptor);
937 398
938 for (; blks; gfs2_replay_incr_blk(jd, !! 399 for (; blks; gfs2_replay_incr_blk(sdp, &start), blks--) {
939 error = gfs2_replay_read_block 400 error = gfs2_replay_read_block(jd, start, &bh);
940 if (error) 401 if (error)
941 return error; 402 return error;
942 403
943 if (!first) 404 if (!first)
944 gfs2_metatype_check(sd 405 gfs2_metatype_check(sdp, bh, GFS2_METATYPE_LB);
945 406
946 while (offset + sizeof(u64) <= 407 while (offset + sizeof(u64) <= sdp->sd_sb.sb_bsize) {
947 blkno = be64_to_cpu(*( 408 blkno = be64_to_cpu(*(__be64 *)(bh->b_data + offset));
948 409
949 error = gfs2_revoke_ad !! 410 error = gfs2_revoke_add(sdp, blkno, start);
950 if (error < 0) { 411 if (error < 0) {
951 brelse(bh); 412 brelse(bh);
952 return error; 413 return error;
953 } 414 }
954 else if (error) 415 else if (error)
955 jd->jd_found_r !! 416 sdp->sd_found_revokes++;
956 417
957 if (!--revokes) 418 if (!--revokes)
958 break; 419 break;
959 offset += sizeof(u64); 420 offset += sizeof(u64);
960 } 421 }
961 422
962 brelse(bh); 423 brelse(bh);
963 offset = sizeof(struct gfs2_me 424 offset = sizeof(struct gfs2_meta_header);
964 first = 0; 425 first = 0;
965 } 426 }
966 427
967 return 0; 428 return 0;
968 } 429 }
969 430
970 static void revoke_lo_after_scan(struct gfs2_j 431 static void revoke_lo_after_scan(struct gfs2_jdesc *jd, int error, int pass)
971 { 432 {
972 struct gfs2_sbd *sdp = GFS2_SB(jd->jd_ 433 struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
973 434
974 if (error) { 435 if (error) {
975 gfs2_revoke_clean(jd); !! 436 gfs2_revoke_clean(sdp);
976 return; 437 return;
977 } 438 }
978 if (pass != 1) 439 if (pass != 1)
979 return; 440 return;
980 441
981 fs_info(sdp, "jid=%u: Found %u revoke 442 fs_info(sdp, "jid=%u: Found %u revoke tags\n",
982 jd->jd_jid, jd->jd_found_revok !! 443 jd->jd_jid, sdp->sd_found_revokes);
983 444
984 gfs2_revoke_clean(jd); !! 445 gfs2_revoke_clean(sdp);
>> 446 }
>> 447
>> 448 static void rg_lo_add(struct gfs2_sbd *sdp, struct gfs2_log_element *le)
>> 449 {
>> 450 struct gfs2_rgrpd *rgd;
>> 451 struct gfs2_trans *tr = current->journal_info;
>> 452
>> 453 tr->tr_touched = 1;
>> 454
>> 455 rgd = container_of(le, struct gfs2_rgrpd, rd_le);
>> 456
>> 457 gfs2_log_lock(sdp);
>> 458 if (!list_empty(&le->le_list)){
>> 459 gfs2_log_unlock(sdp);
>> 460 return;
>> 461 }
>> 462 gfs2_rgrp_bh_hold(rgd);
>> 463 sdp->sd_log_num_rg++;
>> 464 list_add(&le->le_list, &sdp->sd_log_le_rg);
>> 465 gfs2_log_unlock(sdp);
>> 466 }
>> 467
>> 468 static void rg_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_ail *ai)
>> 469 {
>> 470 struct list_head *head = &sdp->sd_log_le_rg;
>> 471 struct gfs2_rgrpd *rgd;
>> 472
>> 473 while (!list_empty(head)) {
>> 474 rgd = list_entry(head->next, struct gfs2_rgrpd, rd_le.le_list);
>> 475 list_del_init(&rgd->rd_le.le_list);
>> 476 sdp->sd_log_num_rg--;
>> 477
>> 478 gfs2_rgrp_repolish_clones(rgd);
>> 479 gfs2_rgrp_bh_put(rgd);
>> 480 }
>> 481 gfs2_assert_warn(sdp, !sdp->sd_log_num_rg);
985 } 482 }
986 483
987 /** 484 /**
988 * databuf_lo_before_commit - Scan the data bu !! 485 * databuf_lo_add - Add a databuf to the transaction.
989 * @sdp: The filesystem !! 486 *
990 * @tr: The system transaction being flushed !! 487 * This is used in two distinct cases:
>> 488 * i) In ordered write mode
>> 489 * We put the data buffer on a list so that we can ensure that its
>> 490 * synced to disk at the right time
>> 491 * ii) In journaled data mode
>> 492 * We need to journal the data block in the same way as metadata in
>> 493 * the functions above. The difference is that here we have a tag
>> 494 * which is two __be64's being the block number (as per meta data)
>> 495 * and a flag which says whether the data block needs escaping or
>> 496 * not. This means we need a new log entry for each 251 or so data
>> 497 * blocks, which isn't an enormous overhead but twice as much as
>> 498 * for normal metadata blocks.
991 */ 499 */
>> 500 static void databuf_lo_add(struct gfs2_sbd *sdp, struct gfs2_log_element *le)
>> 501 {
>> 502 struct gfs2_bufdata *bd = container_of(le, struct gfs2_bufdata, bd_le);
>> 503 struct gfs2_trans *tr = current->journal_info;
>> 504 struct address_space *mapping = bd->bd_bh->b_page->mapping;
>> 505 struct gfs2_inode *ip = GFS2_I(mapping->host);
>> 506
>> 507 lock_buffer(bd->bd_bh);
>> 508 gfs2_log_lock(sdp);
>> 509 if (tr) {
>> 510 if (!list_empty(&bd->bd_list_tr))
>> 511 goto out;
>> 512 tr->tr_touched = 1;
>> 513 if (gfs2_is_jdata(ip)) {
>> 514 tr->tr_num_buf++;
>> 515 list_add(&bd->bd_list_tr, &tr->tr_list_buf);
>> 516 }
>> 517 }
>> 518 if (!list_empty(&le->le_list))
>> 519 goto out;
>> 520
>> 521 set_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags);
>> 522 set_bit(GLF_DIRTY, &bd->bd_gl->gl_flags);
>> 523 if (gfs2_is_jdata(ip)) {
>> 524 gfs2_pin(sdp, bd->bd_bh);
>> 525 tr->tr_num_databuf_new++;
>> 526 sdp->sd_log_num_databuf++;
>> 527 list_add(&le->le_list, &sdp->sd_log_le_databuf);
>> 528 } else {
>> 529 list_add(&le->le_list, &sdp->sd_log_le_ordered);
>> 530 }
>> 531 out:
>> 532 gfs2_log_unlock(sdp);
>> 533 unlock_buffer(bd->bd_bh);
>> 534 }
992 535
993 static void databuf_lo_before_commit(struct gf !! 536 static void gfs2_check_magic(struct buffer_head *bh)
994 { 537 {
995 unsigned int limit = databuf_limit(sdp !! 538 void *kaddr;
996 unsigned int nbuf; !! 539 __be32 *ptr;
997 if (tr == NULL) !! 540
>> 541 clear_buffer_escaped(bh);
>> 542 kaddr = kmap_atomic(bh->b_page, KM_USER0);
>> 543 ptr = kaddr + bh_offset(bh);
>> 544 if (*ptr == cpu_to_be32(GFS2_MAGIC))
>> 545 set_buffer_escaped(bh);
>> 546 kunmap_atomic(kaddr, KM_USER0);
>> 547 }
>> 548
>> 549 static void gfs2_write_blocks(struct gfs2_sbd *sdp, struct buffer_head *bh,
>> 550 struct list_head *list, struct list_head *done,
>> 551 unsigned int n)
>> 552 {
>> 553 struct buffer_head *bh1;
>> 554 struct gfs2_log_descriptor *ld;
>> 555 struct gfs2_bufdata *bd;
>> 556 __be64 *ptr;
>> 557
>> 558 if (!bh)
998 return; 559 return;
999 nbuf = tr->tr_num_databuf_new - tr->tr !! 560
1000 gfs2_before_commit(sdp, limit, nbuf, !! 561 ld = bh_log_desc(bh);
>> 562 ld->ld_length = cpu_to_be32(n + 1);
>> 563 ld->ld_data1 = cpu_to_be32(n);
>> 564
>> 565 ptr = bh_log_ptr(bh);
>> 566
>> 567 get_bh(bh);
>> 568 submit_bh(WRITE_SYNC_PLUG, bh);
>> 569 gfs2_log_lock(sdp);
>> 570 while(!list_empty(list)) {
>> 571 bd = list_entry(list->next, struct gfs2_bufdata, bd_le.le_list);
>> 572 list_move_tail(&bd->bd_le.le_list, done);
>> 573 get_bh(bd->bd_bh);
>> 574 while (be64_to_cpu(*ptr) != bd->bd_bh->b_blocknr) {
>> 575 gfs2_log_incr_head(sdp);
>> 576 ptr += 2;
>> 577 }
>> 578 gfs2_log_unlock(sdp);
>> 579 lock_buffer(bd->bd_bh);
>> 580 if (buffer_escaped(bd->bd_bh)) {
>> 581 void *kaddr;
>> 582 bh1 = gfs2_log_get_buf(sdp);
>> 583 kaddr = kmap_atomic(bd->bd_bh->b_page, KM_USER0);
>> 584 memcpy(bh1->b_data, kaddr + bh_offset(bd->bd_bh),
>> 585 bh1->b_size);
>> 586 kunmap_atomic(kaddr, KM_USER0);
>> 587 *(__be32 *)bh1->b_data = 0;
>> 588 clear_buffer_escaped(bd->bd_bh);
>> 589 unlock_buffer(bd->bd_bh);
>> 590 brelse(bd->bd_bh);
>> 591 } else {
>> 592 bh1 = gfs2_log_fake_buf(sdp, bd->bd_bh);
>> 593 }
>> 594 submit_bh(WRITE_SYNC_PLUG, bh1);
>> 595 gfs2_log_lock(sdp);
>> 596 ptr += 2;
>> 597 }
>> 598 gfs2_log_unlock(sdp);
>> 599 brelse(bh);
>> 600 }
>> 601
>> 602 /**
>> 603 * databuf_lo_before_commit - Scan the data buffers, writing as we go
>> 604 *
>> 605 */
>> 606
>> 607 static void databuf_lo_before_commit(struct gfs2_sbd *sdp)
>> 608 {
>> 609 struct gfs2_bufdata *bd = NULL;
>> 610 struct buffer_head *bh = NULL;
>> 611 unsigned int n = 0;
>> 612 __be64 *ptr = NULL, *end = NULL;
>> 613 LIST_HEAD(processed);
>> 614 LIST_HEAD(in_progress);
>> 615
>> 616 gfs2_log_lock(sdp);
>> 617 while (!list_empty(&sdp->sd_log_le_databuf)) {
>> 618 if (ptr == end) {
>> 619 gfs2_log_unlock(sdp);
>> 620 gfs2_write_blocks(sdp, bh, &in_progress, &processed, n);
>> 621 n = 0;
>> 622 bh = gfs2_get_log_desc(sdp, GFS2_LOG_DESC_JDATA);
>> 623 ptr = bh_log_ptr(bh);
>> 624 end = bh_ptr_end(bh) - 1;
>> 625 gfs2_log_lock(sdp);
>> 626 continue;
>> 627 }
>> 628 bd = list_entry(sdp->sd_log_le_databuf.next, struct gfs2_bufdata, bd_le.le_list);
>> 629 list_move_tail(&bd->bd_le.le_list, &in_progress);
>> 630 gfs2_check_magic(bd->bd_bh);
>> 631 *ptr++ = cpu_to_be64(bd->bd_bh->b_blocknr);
>> 632 *ptr++ = cpu_to_be64(buffer_escaped(bh) ? 1 : 0);
>> 633 n++;
>> 634 }
>> 635 gfs2_log_unlock(sdp);
>> 636 gfs2_write_blocks(sdp, bh, &in_progress, &processed, n);
>> 637 gfs2_log_lock(sdp);
>> 638 list_splice(&processed, &sdp->sd_log_le_databuf);
>> 639 gfs2_log_unlock(sdp);
1001 } 640 }
1002 641
1003 static int databuf_lo_scan_elements(struct gf !! 642 static int databuf_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start,
1004 struct gf 643 struct gfs2_log_descriptor *ld,
1005 __be64 *p 644 __be64 *ptr, int pass)
1006 { 645 {
1007 struct gfs2_inode *ip = GFS2_I(jd->jd 646 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
>> 647 struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
1008 struct gfs2_glock *gl = ip->i_gl; 648 struct gfs2_glock *gl = ip->i_gl;
1009 unsigned int blks = be32_to_cpu(ld->l 649 unsigned int blks = be32_to_cpu(ld->ld_data1);
1010 struct buffer_head *bh_log, *bh_ip; 650 struct buffer_head *bh_log, *bh_ip;
1011 u64 blkno; 651 u64 blkno;
1012 u64 esc; 652 u64 esc;
1013 int error = 0; 653 int error = 0;
1014 654
1015 if (pass != 1 || be32_to_cpu(ld->ld_t 655 if (pass != 1 || be32_to_cpu(ld->ld_type) != GFS2_LOG_DESC_JDATA)
1016 return 0; 656 return 0;
1017 657
1018 gfs2_replay_incr_blk(jd, &start); !! 658 gfs2_replay_incr_blk(sdp, &start);
1019 for (; blks; gfs2_replay_incr_blk(jd, !! 659 for (; blks; gfs2_replay_incr_blk(sdp, &start), blks--) {
1020 blkno = be64_to_cpu(*ptr++); 660 blkno = be64_to_cpu(*ptr++);
1021 esc = be64_to_cpu(*ptr++); 661 esc = be64_to_cpu(*ptr++);
1022 662
1023 jd->jd_found_blocks++; !! 663 sdp->sd_found_blocks++;
1024 664
1025 if (gfs2_revoke_check(jd, blk !! 665 if (gfs2_revoke_check(sdp, blkno, start))
1026 continue; 666 continue;
1027 667
1028 error = gfs2_replay_read_bloc 668 error = gfs2_replay_read_block(jd, start, &bh_log);
1029 if (error) 669 if (error)
1030 return error; 670 return error;
1031 671
1032 bh_ip = gfs2_meta_new(gl, blk 672 bh_ip = gfs2_meta_new(gl, blkno);
1033 memcpy(bh_ip->b_data, bh_log- 673 memcpy(bh_ip->b_data, bh_log->b_data, bh_log->b_size);
1034 674
1035 /* Unescape */ 675 /* Unescape */
1036 if (esc) { 676 if (esc) {
1037 __be32 *eptr = (__be3 677 __be32 *eptr = (__be32 *)bh_ip->b_data;
1038 *eptr = cpu_to_be32(G 678 *eptr = cpu_to_be32(GFS2_MAGIC);
1039 } 679 }
1040 mark_buffer_dirty(bh_ip); 680 mark_buffer_dirty(bh_ip);
1041 681
1042 brelse(bh_log); 682 brelse(bh_log);
1043 brelse(bh_ip); 683 brelse(bh_ip);
>> 684 if (error)
>> 685 break;
1044 686
1045 jd->jd_replayed_blocks++; !! 687 sdp->sd_replayed_blocks++;
1046 } 688 }
1047 689
1048 return error; 690 return error;
1049 } 691 }
1050 692
1051 /* FIXME: sort out accounting for log blocks 693 /* FIXME: sort out accounting for log blocks etc. */
1052 694
1053 static void databuf_lo_after_scan(struct gfs2 695 static void databuf_lo_after_scan(struct gfs2_jdesc *jd, int error, int pass)
1054 { 696 {
1055 struct gfs2_inode *ip = GFS2_I(jd->jd 697 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
1056 struct gfs2_sbd *sdp = GFS2_SB(jd->jd 698 struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
1057 699
1058 if (error) { 700 if (error) {
1059 gfs2_inode_metasync(ip->i_gl) !! 701 gfs2_meta_sync(ip->i_gl);
1060 return; 702 return;
1061 } 703 }
1062 if (pass != 1) 704 if (pass != 1)
1063 return; 705 return;
1064 706
1065 /* data sync? */ 707 /* data sync? */
1066 gfs2_inode_metasync(ip->i_gl); !! 708 gfs2_meta_sync(ip->i_gl);
1067 709
1068 fs_info(sdp, "jid=%u: Replayed %u of 710 fs_info(sdp, "jid=%u: Replayed %u of %u data blocks\n",
1069 jd->jd_jid, jd->jd_replayed_b !! 711 jd->jd_jid, sdp->sd_replayed_blocks, sdp->sd_found_blocks);
1070 } 712 }
1071 713
1072 static void databuf_lo_after_commit(struct gf !! 714 static void databuf_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_ail *ai)
1073 { 715 {
1074 struct list_head *head; !! 716 struct list_head *head = &sdp->sd_log_le_databuf;
1075 struct gfs2_bufdata *bd; 717 struct gfs2_bufdata *bd;
1076 718
1077 if (tr == NULL) <<
1078 return; <<
1079 <<
1080 head = &tr->tr_databuf; <<
1081 while (!list_empty(head)) { 719 while (!list_empty(head)) {
1082 bd = list_first_entry(head, s !! 720 bd = list_entry(head->next, struct gfs2_bufdata, bd_le.le_list);
1083 list_del_init(&bd->bd_list); !! 721 list_del_init(&bd->bd_le.le_list);
1084 gfs2_unpin(sdp, bd->bd_bh, tr !! 722 sdp->sd_log_num_databuf--;
>> 723 gfs2_unpin(sdp, bd->bd_bh, ai);
1085 } 724 }
>> 725 gfs2_assert_warn(sdp, !sdp->sd_log_num_databuf);
1086 } 726 }
1087 727
1088 728
1089 static const struct gfs2_log_operations gfs2_ !! 729 const struct gfs2_log_operations gfs2_buf_lops = {
>> 730 .lo_add = buf_lo_add,
1090 .lo_before_commit = buf_lo_before_com 731 .lo_before_commit = buf_lo_before_commit,
1091 .lo_after_commit = buf_lo_after_commi 732 .lo_after_commit = buf_lo_after_commit,
1092 .lo_before_scan = buf_lo_before_scan, 733 .lo_before_scan = buf_lo_before_scan,
1093 .lo_scan_elements = buf_lo_scan_eleme 734 .lo_scan_elements = buf_lo_scan_elements,
1094 .lo_after_scan = buf_lo_after_scan, 735 .lo_after_scan = buf_lo_after_scan,
1095 .lo_name = "buf", 736 .lo_name = "buf",
1096 }; 737 };
1097 738
1098 static const struct gfs2_log_operations gfs2_ !! 739 const struct gfs2_log_operations gfs2_revoke_lops = {
>> 740 .lo_add = revoke_lo_add,
1099 .lo_before_commit = revoke_lo_before_ 741 .lo_before_commit = revoke_lo_before_commit,
1100 .lo_after_commit = revoke_lo_after_co <<
1101 .lo_before_scan = revoke_lo_before_sc 742 .lo_before_scan = revoke_lo_before_scan,
1102 .lo_scan_elements = revoke_lo_scan_el 743 .lo_scan_elements = revoke_lo_scan_elements,
1103 .lo_after_scan = revoke_lo_after_scan 744 .lo_after_scan = revoke_lo_after_scan,
1104 .lo_name = "revoke", 745 .lo_name = "revoke",
1105 }; 746 };
1106 747
1107 static const struct gfs2_log_operations gfs2_ !! 748 const struct gfs2_log_operations gfs2_rg_lops = {
>> 749 .lo_add = rg_lo_add,
>> 750 .lo_after_commit = rg_lo_after_commit,
>> 751 .lo_name = "rg",
>> 752 };
>> 753
>> 754 const struct gfs2_log_operations gfs2_databuf_lops = {
>> 755 .lo_add = databuf_lo_add,
1108 .lo_before_commit = databuf_lo_before 756 .lo_before_commit = databuf_lo_before_commit,
1109 .lo_after_commit = databuf_lo_after_c 757 .lo_after_commit = databuf_lo_after_commit,
1110 .lo_scan_elements = databuf_lo_scan_e 758 .lo_scan_elements = databuf_lo_scan_elements,
1111 .lo_after_scan = databuf_lo_after_sca 759 .lo_after_scan = databuf_lo_after_scan,
1112 .lo_name = "databuf", 760 .lo_name = "databuf",
1113 }; 761 };
1114 762
1115 const struct gfs2_log_operations *gfs2_log_op 763 const struct gfs2_log_operations *gfs2_log_ops[] = {
1116 &gfs2_databuf_lops, 764 &gfs2_databuf_lops,
1117 &gfs2_buf_lops, 765 &gfs2_buf_lops,
>> 766 &gfs2_rg_lops,
1118 &gfs2_revoke_lops, 767 &gfs2_revoke_lops,
1119 NULL, 768 NULL,
1120 }; 769 };
1121 770
1122 771
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.