aboutsummaryrefslogtreecommitdiffstats
path: root/fs/ocfs2/aops.c
blob: acf8f0006725e26c7a58f922e69e1c3ffde2e4d6 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
/* -*- mode: c; c-basic-offset: 8; -*-
 * vim: noexpandtab sw=8 ts=8 sts=0:
 *
 * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License as published by the Free Software Foundation; either
 * version 2 of the License, or (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

#include <linux/fs.h>
#include <linux/slab.h>
#include <linux/highmem.h>
#include <linux/pagemap.h>
#include <asm/byteorder.h>
#include <linux/swap.h>

#define MLOG_MASK_PREFIX ML_FILE_IO
#include <cluster/masklog.h>

#include "ocfs2.h"

#include "alloc.h"
#include "aops.h"
#include "dlmglue.h"
#include "extent_map.h"
#include "file.h"
#include "inode.h"
#include "journal.h"
#include "suballoc.h"
#include "super.h"
#include "symlink.h"

#include "buffer_head_io.h"

static int ocfs2_symlink_get_block(struct inode *inode, sector_t iblock,
				   struct buffer_head *bh_result, int create)
{
	int err = -EIO;
	int status;
	struct ocfs2_dinode *fe = NULL;
	struct buffer_head *bh = NULL;
	struct buffer_head *buffer_cache_bh = NULL;
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	void *kaddr;

	mlog_entry("(0x%p, %llu, 0x%p, %d)\n", inode,
		   (unsigned long long)iblock, bh_result, create);

	BUG_ON(ocfs2_inode_is_fast_symlink(inode));

	if ((iblock << inode->i_sb->s_blocksize_bits) > PATH_MAX + 1) {
		mlog(ML_ERROR, "block offset > PATH_MAX: %llu",
		     (unsigned long long)iblock);
		goto bail;
	}

	status = ocfs2_read_block(OCFS2_SB(inode->i_sb),
				  OCFS2_I(inode)->ip_blkno,
				  &bh, OCFS2_BH_CACHED, inode);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}
	fe = (struct ocfs2_dinode *) bh->b_data;

	if (!OCFS2_IS_VALID_DINODE(fe)) {
		mlog(ML_ERROR, "Invalid dinode #%llu: signature = %.*s\n",
		     (unsigned long long)fe->i_blkno, 7, fe->i_signature);
		goto bail;
	}

	if ((u64)iblock >= ocfs2_clusters_to_blocks(inode->i_sb,
						    le32_to_cpu(fe->i_clusters))) {
		mlog(ML_ERROR, "block offset is outside the allocated size: "
		     "%llu\n", (unsigned long long)iblock);
		goto bail;
	}

	/* We don't use the page cache to create symlink data, so if
	 * need be, copy it over from the buffer cache. */
	if (!buffer_uptodate(bh_result) && ocfs2_inode_is_new(inode)) {
		u64 blkno = le64_to_cpu(fe->id2.i_list.l_recs[0].e_blkno) +
			    iblock;
		buffer_cache_bh = sb_getblk(osb->sb, blkno);
		if (!buffer_cache_bh) {
			mlog(ML_ERROR, "couldn't getblock for symlink!\n");
			goto bail;
		}

		/* we haven't locked out transactions, so a commit
		 * could've happened. Since we've got a reference on
		 * the bh, even if it commits while we're doing the
		 * copy, the data is still good. */
		if (buffer_jbd(buffer_cache_bh)
		    && ocfs2_inode_is_new(inode)) {
			kaddr = kmap_atomic(bh_result->b_page, KM_USER0);
			if (!kaddr) {
				mlog(ML_ERROR, "couldn't kmap!\n");
				goto bail;
			}
			memcpy(kaddr + (bh_result->b_size * iblock),
			       buffer_cache_bh->b_data,
			       bh_result->b_size);
			kunmap_atomic(kaddr, KM_USER0);
			set_buffer_uptodate(bh_result);
		}
		brelse(buffer_cache_bh);
	}

	map_bh(bh_result, inode->i_sb,
	       le64_to_cpu(fe->id2.i_list.l_recs[0].e_blkno) + iblock);

	err = 0;

bail:
	if (bh)
		brelse(bh);

	mlog_exit(err);
	return err;
}

static int ocfs2_get_block(struct inode *inode, sector_t iblock,
			   struct buffer_head *bh_result, int create)
{
	int err = 0;
	u64 p_blkno, past_eof;
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);

	mlog_entry("(0x%p, %llu, 0x%p, %d)\n", inode,
		   (unsigned long long)iblock, bh_result, create);

	if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SYSTEM_FILE)
		mlog(ML_NOTICE, "get_block on system inode 0x%p (%lu)\n",
		     inode, inode->i_ino);

	if (S_ISLNK(inode->i_mode)) {
		/* this always does I/O for some reason. */
		err = ocfs2_symlink_get_block(inode, iblock, bh_result, create);
		goto bail;
	}

	err = ocfs2_extent_map_get_blocks(inode, iblock, &p_blkno, NULL);
	if (err) {
		mlog(ML_ERROR, "Error %d from get_blocks(0x%p, %llu, 1, "
		     "%llu, NULL)\n", err, inode, (unsigned long long)iblock,
		     (unsigned long long)p_blkno);
		goto bail;
	}

	/*
	 * ocfs2 never allocates in this function - the only time we
	 * need to use BH_New is when we're extending i_size on a file
	 * system which doesn't support holes, in which case BH_New
	 * allows block_prepare_write() to zero.
	 */
	mlog_bug_on_msg(create && p_blkno == 0 && ocfs2_sparse_alloc(osb),
			"ino %lu, iblock %llu\n", inode->i_ino,
			(unsigned long long)iblock);

	if (p_blkno)
		map_bh(bh_result, inode->i_sb, p_blkno);

	if (!ocfs2_sparse_alloc(osb)) {
		if (p_blkno == 0) {
			err = -EIO;
			mlog(ML_ERROR,
			     "iblock = %llu p_blkno = %llu blkno=(%llu)\n",
			     (unsigned long long)iblock,
			     (unsigned long long)p_blkno,
			     (unsigned long long)OCFS2_I(inode)->ip_blkno);
			mlog(ML_ERROR, "Size %llu, clusters %u\n", (unsigned long long)i_size_read(inode), OCFS2_I(inode)->ip_clusters);
			dump_stack();
		}

		past_eof = ocfs2_blocks_for_bytes(inode->i_sb, i_size_read(inode));
		mlog(0, "Inode %lu, past_eof = %llu\n", inode->i_ino,
		     (unsigned long long)past_eof);

		if (create && (iblock >= past_eof))
			set_buffer_new(bh_result);
	}

bail:
	if (err < 0)
		err = -EIO;

	mlog_exit(err);
	return err;
}

static int ocfs2_readpage(struct file *file, struct page *page)
{
	struct inode *inode = page->mapping->host;
	loff_t start = (loff_t)page->index << PAGE_CACHE_SHIFT;
	int ret, unlock = 1;

	mlog_entry("(0x%p, %lu)\n", file, (page ? page->index : 0));

	ret = ocfs2_meta_lock_with_page(inode, NULL, 0, page);
	if (ret != 0) {
		if (ret == AOP_TRUNCATED_PAGE)
			unlock = 0;
		mlog_errno(ret);
		goto out;
	}

	down_read(&OCFS2_I(inode)->ip_alloc_sem);

	/*
	 * i_size might have just been updated as we grabed the meta lock.  We
	 * might now be discovering a truncate that hit on another node.
	 * block_read_full_page->get_block freaks out if it is asked to read
	 * beyond the end of a file, so we check here.  Callers
	 * (generic_file_read, fault->nopage) are clever enough to check i_size
	 * and notice that the page they just read isn't needed.
	 *
	 * XXX sys_readahead() seems to get that wrong?
	 */
	if (start >= i_size_read(inode)) {
		char *addr = kmap(page);
		memset(addr, 0, PAGE_SIZE);
		flush_dcache_page(page);
		kunmap(page);
		SetPageUptodate(page);
		ret = 0;
		goto out_alloc;
	}

	ret = ocfs2_data_lock_with_page(inode, 0, page);
	if (ret != 0) {
		if (ret == AOP_TRUNCATED_PAGE)
			unlock = 0;
		mlog_errno(ret);
		goto out_alloc;
	}

	ret = block_read_full_page(page, ocfs2_get_block);
	unlock = 0;

	ocfs2_data_unlock(inode, 0);
out_alloc:
	up_read(&OCFS2_I(inode)->ip_alloc_sem);
	ocfs2_meta_unlock(inode, 0);
out:
	if (unlock)
		unlock_page(page);
	mlog_exit(ret);
	return ret;
}

/* Note: Because we don't support holes, our allocation has
 * already happened (allocation writes zeros to the file data)
 * so we don't have to worry about ordered writes in
 * ocfs2_writepage.
 *
 * ->writepage is called during the process of invalidating the page cache
 * during blocked lock processing.  It can't block on any cluster locks
 * to during block mapping.  It's relying on the fact that the block
 * mapping can't have disappeared under the dirty pages that it is
 * being asked to write back.
 */
static int ocfs2_writepage(struct page *page, struct writeback_control *wbc)
{
	int ret;

	mlog_entry("(0x%p)\n", page);

	ret = block_write_full_page(page, ocfs2_get_block, wbc);

	mlog_exit(ret);

	return ret;
}

/*
 * This is called from ocfs2_write_zero_page() which has handled it's
 * own cluster locking and has ensured allocation exists for those
 * blocks to be written.
 */
int ocfs2_prepare_write_nolock(struct inode *inode, struct page *page,
			       unsigned from, unsigned to)
{
	int ret;

	down_read(&OCFS2_I(inode)->ip_alloc_sem);

	ret = block_prepare_write(page, from, to, ocfs2_get_block);

	up_read(&OCFS2_I(inode)->ip_alloc_sem);

	return ret;
}

/* Taken from ext3. We don't necessarily need the full blown
 * functionality yet, but IMHO it's better to cut and paste the whole
 * thing so we can avoid introducing our own bugs (and easily pick up
 * their fixes when they happen) --Mark */
static int walk_page_buffers(	handle_t *handle,
				struct buffer_head *head,
				unsigned from,
				unsigned to,
				int *partial,
				int (*fn)(	handle_t *handle,
						struct buffer_head *bh))
{
	struct buffer_head *bh;
	unsigned block_start, block_end;
	unsigned blocksize = head->b_size;
	int err, ret = 0;
	struct buffer_head *next;

	for (	bh = head, block_start = 0;
		ret == 0 && (bh != head || !block_start);
	    	block_start = block_end, bh = next)
	{
		next = bh->b_this_page;
		block_end = block_start + blocksize;
		if (block_end <= from || block_start >= to) {
			if (partial && !buffer_uptodate(bh))
				*partial = 1;
			continue;
		}
		err = (*fn)(handle, bh);
		if (!ret)
			ret = err;
	}
	return ret;
}

handle_t *ocfs2_start_walk_page_trans(struct inode *inode,
							 struct page *page,
							 unsigned from,
							 unsigned to)
{
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	handle_t *handle = NULL;
	int ret = 0;

	handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
	if (!handle) {
		ret = -ENOMEM;
		mlog_errno(ret);
		goto out;
	}

	if (ocfs2_should_order_data(inode)) {
		ret = walk_page_buffers(handle,
					page_buffers(page),
					from, to, NULL,
					ocfs2_journal_dirty_data);
		if (ret < 0) 
			mlog_errno(ret);
	}
out:
	if (ret) {
		if (handle)
			ocfs2_commit_trans(osb, handle);
		handle = ERR_PTR(ret);
	}
	return handle;
}

static sector_t ocfs2_bmap(struct address_space *mapping, sector_t block)
{
	sector_t status;
	u64 p_blkno = 0;
	int err = 0;
	struct inode *inode = mapping->host;

	mlog_entry("(block = %llu)\n", (unsigned long long)block);

	/* We don't need to lock journal system files, since they aren't
	 * accessed concurrently from multiple nodes.
	 */
	if (!INODE_JOURNAL(inode)) {
		err = ocfs2_meta_lock(inode, NULL, 0);
		if (err) {
			if (err != -ENOENT)
				mlog_errno(err);
			goto bail;
		}
		down_read(&OCFS2_I(inode)->ip_alloc_sem);
	}

	err = ocfs2_extent_map_get_blocks(inode, block, &p_blkno, NULL);

	if (!INODE_JOURNAL(inode)) {
		up_read(&OCFS2_I(inode)->ip_alloc_sem);
		ocfs2_meta_unlock(inode, 0);
	}

	if (err) {
		mlog(ML_ERROR, "get_blocks() failed, block = %llu\n",
		     (unsigned long long)block);
		mlog_errno(err);
		goto bail;
	}


bail:
	status = err ? 0 : p_blkno;

	mlog_exit((int)status);

	return status;
}

/*
 * TODO: Make this into a generic get_blocks function.
 *
 * From do_direct_io in direct-io.c:
 *  "So what we do is to permit the ->get_blocks function to populate
 *   bh.b_size with the size of IO which is permitted at this offset and
 *   this i_blkbits."
 *
 * This function is called directly from get_more_blocks in direct-io.c.
 *
 * called like this: dio->get_blocks(dio->inode, fs_startblk,
 * 					fs_count, map_bh, dio->rw == WRITE);
 */
static int ocfs2_direct_IO_get_blocks(struct inode *inode, sector_t iblock,
				     struct buffer_head *bh_result, int create)
{
	int ret;
	u64 p_blkno, inode_blocks;
	int contig_blocks;
	unsigned char blocksize_bits = inode->i_sb->s_blocksize_bits;
	unsigned long max_blocks = bh_result->b_size >> inode->i_blkbits;

	/* This function won't even be called if the request isn't all
	 * nicely aligned and of the right size, so there's no need
	 * for us to check any of that. */

	inode_blocks = ocfs2_blocks_for_bytes(inode->i_sb, i_size_read(inode));

	/*
	 * Any write past EOF is not allowed because we'd be extending.
	 */
	if (create && (iblock + max_blocks) > inode_blocks) {
		ret = -EIO;
		goto bail;
	}

	/* This figures out the size of the next contiguous block, and
	 * our logical offset */
	ret = ocfs2_extent_map_get_blocks(inode, iblock, &p_blkno,
					  &contig_blocks);
	if (ret) {
		mlog(ML_ERROR, "get_blocks() failed iblock=%llu\n",
		     (unsigned long long)iblock);
		ret = -EIO;
		goto bail;
	}

	if (!ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb)) && !p_blkno) {
		ocfs2_error(inode->i_sb,
			    "Inode %llu has a hole at block %llu\n",
			    (unsigned long long)OCFS2_I(inode)->ip_blkno,
			    (unsigned long long)iblock);
		ret = -EROFS;
		goto bail;
	}

	/*
	 * get_more_blocks() expects us to describe a hole by clearing
	 * the mapped bit on bh_result().
	 */
	if (p_blkno)
		map_bh(bh_result, inode->i_sb, p_blkno);
	else {
		/*
		 * ocfs2_prepare_inode_for_write() should have caught
		 * the case where we'd be filling a hole and triggered
		 * a buffered write instead.
		 */
		if (create) {
			ret = -EIO;
			mlog_errno(ret);
			goto bail;
		}

		clear_buffer_mapped(bh_result);
	}

	/* make sure we don't map more than max_blocks blocks here as
	   that's all the kernel will handle at this point. */
	if (max_blocks < contig_blocks)
		contig_blocks = max_blocks;
	bh_result->b_size = contig_blocks << blocksize_bits;
bail:
	return ret;
}

/* 
 * ocfs2_dio_end_io is called by the dio core when a dio is finished.  We're
 * particularly interested in the aio/dio case.  Like the core uses
 * i_alloc_sem, we use the rw_lock DLM lock to protect io on one node from
 * truncation on another.
 */
static void ocfs2_dio_end_io(struct kiocb *iocb,
			     loff_t offset,
			     ssize_t bytes,
			     void *private)
{
	struct inode *inode = iocb->ki_filp->f_path.dentry->d_inode;

	/* this io's submitter should not have unlocked this before we could */
	BUG_ON(!ocfs2_iocb_is_rw_locked(iocb));
	ocfs2_iocb_clear_rw_locked(iocb);
	up_read(&inode->i_alloc_sem);
	ocfs2_rw_unlock(inode, 0);
}

/*
 * ocfs2_invalidatepage() and ocfs2_releasepage() are shamelessly stolen
 * from ext3.  PageChecked() bits have been removed as OCFS2 does not
 * do journalled data.
 */
static void ocfs2_invalidatepage(struct page *page, unsigned long offset)
{
	journal_t *journal = OCFS2_SB(page->mapping->host->i_sb)->journal->j_journal;

	journal_invalidatepage(journal, page, offset);
}

static int ocfs2_releasepage(struct page *page, gfp_t wait)
{
	journal_t *journal = OCFS2_SB(page->mapping->host->i_sb)->journal->j_journal;

	if (!page_has_buffers(page))
		return 0;
	return journal_try_to_free_buffers(journal, page, wait);
}

static ssize_t ocfs2_direct_IO(int rw,
			       struct kiocb *iocb,
			       const struct iovec *iov,
			       loff_t offset,
			       unsigned long nr_segs)
{
	struct file *file = iocb->ki_filp;
	struct inode *inode = file->f_path.dentry->d_inode->i_mapping->host;
	int ret;

	mlog_entry_void();

	if (!ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) {
		/*
		 * We get PR data locks even for O_DIRECT.  This
		 * allows concurrent O_DIRECT I/O but doesn't let
		 * O_DIRECT with extending and buffered zeroing writes
		 * race.  If they did race then the buffered zeroing
		 * could be written back after the O_DIRECT I/O.  It's
		 * one thing to tell people not to mix buffered and
		 * O_DIRECT writes, but expecting them to understand
		 * that file extension is also an implicit buffered
		 * write is too much.  By getting the PR we force
		 * writeback of the buffered zeroing before
		 * proceeding.
		 */
		ret = ocfs2_data_lock(inode, 0);
		if (ret < 0) {
			mlog_errno(ret);
			goto out;
		}
		ocfs2_data_unlock(inode, 0);
	}

	ret = blockdev_direct_IO_no_locking(rw, iocb, inode,
					    inode->i_sb->s_bdev, iov, offset,
					    nr_segs, 
					    ocfs2_direct_IO_get_blocks,
					    ocfs2_dio_end_io);
out:
	mlog_exit(ret);
	return ret;
}

static void ocfs2_figure_cluster_boundaries(struct ocfs2_super *osb,
					    u32 cpos,
					    unsigned int *start,
					    unsigned int *end)
{
	unsigned int cluster_start = 0, cluster_end = PAGE_CACHE_SIZE;

	if (unlikely(PAGE_CACHE_SHIFT > osb->s_clustersize_bits)) {
		unsigned int cpp;

		cpp = 1 << (PAGE_CACHE_SHIFT - osb->s_clustersize_bits);

		cluster_start = cpos % cpp;
		cluster_start = cluster_start << osb->s_clustersize_bits;

		cluster_end = cluster_start + osb->s_clustersize;
	}

	BUG_ON(cluster_start > PAGE_SIZE);
	BUG_ON(cluster_end > PAGE_SIZE);

	if (start)
		*start = cluster_start;
	if (end)
		*end = cluster_end;
}

/*
 * 'from' and 'to' are the region in the page to avoid zeroing.
 *
 * If pagesize > clustersize, this function will avoid zeroing outside
 * of the cluster boundary.
 *
 * from == to == 0 is code for "zero the entire cluster region"
 */
static void ocfs2_clear_page_regions(struct page *page,
				     struct ocfs2_super *osb, u32 cpos,
				     unsigned from, unsigned to)
{
	void *kaddr;
	unsigned int cluster_start, cluster_end;

	ocfs2_figure_cluster_boundaries(osb, cpos, &cluster_start, &cluster_end);

	kaddr = kmap_atomic(page, KM_USER0);

	if (from || to) {
		if (from > cluster_start)
			memset(kaddr + cluster_start, 0, from - cluster_start);
		if (to < cluster_end)
			memset(kaddr + to, 0, cluster_end - to);
	} else {
		memset(kaddr + cluster_start, 0, cluster_end - cluster_start);
	}

	kunmap_atomic(kaddr, KM_USER0);
}

/*
 * Some of this taken from block_prepare_write(). We already have our
 * mapping by now though, and the entire write will be allocating or
 * it won't, so not much need to use BH_New.
 *
 * This will also skip zeroing, which is handled externally.
 */
static int ocfs2_map_page_blocks(struct page *page, u64 *p_blkno,
				 struct inode *inode, unsigned int from,
				 unsigned int to, int new)
{
	int ret = 0;
	struct buffer_head *head, *bh, *wait[2], **wait_bh = wait;
	unsigned int block_end, block_start;
	unsigned int bsize = 1 << inode->i_blkbits;

	if (!page_has_buffers(page))
		create_empty_buffers(page, bsize, 0);

	head = page_buffers(page);
	for (bh = head, block_start = 0; bh != head || !block_start;
	     bh = bh->b_this_page, block_start += bsize) {
		block_end = block_start + bsize;

		/*
		 * Ignore blocks outside of our i/o range -
		 * they may belong to unallocated clusters.
		 */
		if (block_start >= to ||
		    (block_start + bsize) <= from) {
			if (PageUptodate(page))
				set_buffer_uptodate(bh);
			continue;
		}

		/*
		 * For an allocating write with cluster size >= page
		 * size, we always write the entire page.
		 */

		if (buffer_new(bh))
			clear_buffer_new(bh);

		if (!buffer_mapped(bh)) {
			map_bh(bh, inode->i_sb, *p_blkno);
			unmap_underlying_metadata(bh->b_bdev, bh->b_blocknr);
		}

		if (PageUptodate(page)) {
			if (!buffer_uptodate(bh))
				set_buffer_uptodate(bh);
		} else if (!buffer_uptodate(bh) && !buffer_delay(bh) &&
		     (block_start < from || block_end > to)) {
			ll_rw_block(READ, 1, &bh);
			*wait_bh++=bh;
		}

		*p_blkno = *p_blkno + 1;
	}

	/*
	 * If we issued read requests - let them complete.
	 */
	while(wait_bh > wait) {
		wait_on_buffer(*--wait_bh);
		if (!buffer_uptodate(*wait_bh))
			ret = -EIO;
	}

	if (ret == 0 || !new)
		return ret;

	/*
	 * If we get -EIO above, zero out any newly allocated blocks
	 * to avoid exposing stale data.
	 */
	bh = head;
	block_start = 0;
	do {
		void *kaddr;

		block_end = block_start + bsize;
		if (block_end <= from)
			goto next_bh;
		if (block_start >= to)
			break;

		kaddr = kmap_atomic(page, KM_USER0);
		memset(kaddr+block_start, 0, bh->b_size);
		flush_dcache_page(page);
		kunmap_atomic(kaddr, KM_USER0);
		set_buffer_uptodate(bh);
		mark_buffer_dirty(bh);

next_bh:
		block_start = block_end;
		bh = bh->b_this_page;
	} while (bh != head);

	return ret;
}

/*
 * This will copy user data from the iovec in the buffered write
 * context.
 */
int ocfs2_map_and_write_user_data(struct inode *inode,
				  struct ocfs2_write_ctxt *wc, u64 *p_blkno,
				  unsigned int *ret_from, unsigned int *ret_to)
{
	int ret;
	unsigned int to, from, cluster_start, cluster_end;
	unsigned long bytes, src_from;
	char *dst;
	struct ocfs2_buffered_write_priv *bp = wc->w_private;
	const struct iovec *cur_iov = bp->b_cur_iov;
	char __user *buf;
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);

	ocfs2_figure_cluster_boundaries(osb, wc->w_cpos, &cluster_start,
					&cluster_end);

	buf = cur_iov->iov_base + bp->b_cur_off;
	src_from = (unsigned long)buf & ~PAGE_CACHE_MASK;

	from = wc->w_pos & (PAGE_CACHE_SIZE - 1);

	/*
	 * This is a lot of comparisons, but it reads quite
	 * easily, which is important here.
	 */
	/* Stay within the src page */
	bytes = PAGE_SIZE - src_from;
	/* Stay within the vector */
	bytes = min(bytes,
		    (unsigned long)(cur_iov->iov_len - bp->b_cur_off));
	/* Stay within count */
	bytes = min(bytes, (unsigned long)wc->w_count);
	/*
	 * For clustersize > page size, just stay within
	 * target page, otherwise we have to calculate pos
	 * within the cluster and obey the rightmost
	 * boundary.
	 */
	if (wc->w_large_pages) {
		/*
		 * For cluster size < page size, we have to
		 * calculate pos within the cluster and obey
		 * the rightmost boundary.
		 */
		bytes = min(bytes, (unsigned long)(osb->s_clustersize
				   - (wc->w_pos & (osb->s_clustersize - 1))));
	} else {
		/*
		 * cluster size > page size is the most common
		 * case - we just stay within the target page
		 * boundary.
		 */
		bytes = min(bytes, PAGE_CACHE_SIZE - from);
	}

	to = from + bytes;

	if (wc->w_this_page_new)
		ret = ocfs2_map_page_blocks(wc->w_this_page, p_blkno, inode,
					    cluster_start, cluster_end, 1);
	else
		ret = ocfs2_map_page_blocks(wc->w_this_page, p_blkno, inode,
					    from, to, 0);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	BUG_ON(from > PAGE_CACHE_SIZE);
	BUG_ON(to > PAGE_CACHE_SIZE);
	BUG_ON(from > osb->s_clustersize);
	BUG_ON(to > osb->s_clustersize);

	dst = kmap(wc->w_this_page);
	memcpy(dst + from, bp->b_src_buf + src_from, bytes);
	kunmap(wc->w_this_page);

	/*
	 * XXX: This is slow, but simple. The caller of
	 * ocfs2_buffered_write_cluster() is responsible for
	 * passing through the iovecs, so it's difficult to
	 * predict what our next step is in here after our
	 * initial write. A future version should be pushing
	 * that iovec manipulation further down.
	 *
	 * By setting this, we indicate that a copy from user
	 * data was done, and subsequent calls for this
	 * cluster will skip copying more data.
	 */
	wc->w_finished_copy = 1;

	*ret_from = from;
	*ret_to = to;
out:

	return bytes ? (unsigned int)bytes : ret;
}

/*
 * Map, fill and write a page to disk.
 *
 * The work of copying data is done via callback.  Newly allocated
 * pages which don't take user data will be zero'd (set 'new' to
 * indicate an allocating write)
 *
 * Returns a negative error code or the number of bytes copied into
 * the page.
 */
int ocfs2_write_data_page(struct inode *inode, handle_t *handle,
			  u64 *p_blkno, struct page *page,
			  struct ocfs2_write_ctxt *wc, int new)
{
	int ret, copied = 0;
	unsigned int from = 0, to = 0;
	unsigned int cluster_start, cluster_end;
	unsigned int zero_from = 0, zero_to = 0;

	ocfs2_figure_cluster_boundaries(OCFS2_SB(inode->i_sb), wc->w_cpos,
					&cluster_start, &cluster_end);

	if ((wc->w_pos >> PAGE_CACHE_SHIFT) == page->index
	    && !wc->w_finished_copy) {

		wc->w_this_page = page;
		wc->w_this_page_new = new;
		ret = wc->w_write_data_page(inode, wc, p_blkno, &from, &to);
		if (ret < 0) {
			mlog_errno(ret);
			goto out;
		}

		copied = ret;

		zero_from = from;
		zero_to = to;
		if (new) {
			from = cluster_start;
			to = cluster_end;
		}
	} else {
		/*
		 * If we haven't allocated the new page yet, we
		 * shouldn't be writing it out without copying user
		 * data. This is likely a math error from the caller.
		 */
		BUG_ON(!new);

		from = cluster_start;
		to = cluster_end;

		ret = ocfs2_map_page_blocks(page, p_blkno, inode,
					    cluster_start, cluster_end, 1);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}
	}

	/*
	 * Parts of newly allocated pages need to be zero'd.
	 *
	 * Above, we have also rewritten 'to' and 'from' - as far as
	 * the rest of the function is concerned, the entire cluster
	 * range inside of a page needs to be written.
	 *
	 * We can skip this if the page is up to date - it's already
	 * been zero'd from being read in as a hole.
	 */
	if (new && !PageUptodate(page))
		ocfs2_clear_page_regions(page, OCFS2_SB(inode->i_sb),
					 wc->w_cpos, zero_from, zero_to);

	flush_dcache_page(page);

	if (ocfs2_should_order_data(inode)) {
		ret = walk_page_buffers(handle,
					page_buffers(page),
					from, to, NULL,
					ocfs2_journal_dirty_data);
		if (ret < 0)
			mlog_errno(ret);
	}

	/*
	 * We don't use generic_commit_write() because we need to
	 * handle our own i_size update.
	 */
	ret = block_commit_write(page, from, to);
	if (ret)
		mlog_errno(ret);
out:

	return copied ? copied : ret;
}

/*
 * Do the actual write of some data into an inode. Optionally allocate
 * in order to fulfill the write.
 *
 * cpos is the logical cluster offset within the file to write at
 *
 * 'phys' is the physical mapping of that offset. a 'phys' value of
 * zero indicates that allocation is required. In this case, data_ac
 * and meta_ac should be valid (meta_ac can be null if metadata
 * allocation isn't required).
 */
static ssize_t ocfs2_write(struct file *file, u32 phys, handle_t *handle,
			   struct buffer_head *di_bh,
			   struct ocfs2_alloc_context *data_ac,
			   struct ocfs2_alloc_context *meta_ac,
			   struct ocfs2_write_ctxt *wc)
{
	int ret, i, numpages = 1, new;
	unsigned int copied = 0;
	u32 tmp_pos;
	u64 v_blkno, p_blkno;
	struct address_space *mapping = file->f_mapping;
	struct inode *inode = mapping->host;
	unsigned int cbits = OCFS2_SB(inode->i_sb)->s_clustersize_bits;
	unsigned long index, start;
	struct page **cpages;

	new = phys == 0 ? 1 : 0;

	/*
	 * Figure out how many pages we'll be manipulating here. For
	 * non-allocating write, or any writes where cluster size is
	 * less than page size, we only need one page. Otherwise,
	 * allocating writes of cluster size larger than page size
	 * need cluster size pages.
	 */
	if (new && !wc->w_large_pages)
		numpages = (1 << cbits) / PAGE_SIZE;

	cpages = kzalloc(sizeof(*cpages) * numpages, GFP_NOFS);
	if (!cpages) {
		ret = -ENOMEM;
		mlog_errno(ret);
		return ret;
	}

	/*
	 * Fill our page array first. That way we've grabbed enough so
	 * that we can zero and flush if we error after adding the
	 * extent.
	 */
	if (new) {
		start = ocfs2_align_clusters_to_page_index(inode->i_sb,
							   wc->w_cpos);
		v_blkno = ocfs2_clusters_to_blocks(inode->i_sb, wc->w_cpos);
	} else {
		start = wc->w_pos >> PAGE_CACHE_SHIFT;
		v_blkno = wc->w_pos >> inode->i_sb->s_blocksize_bits;
	}

	for(i = 0; i < numpages; i++) {
		index = start + i;

		cpages[i] = grab_cache_page(mapping, index);
		if (!cpages[i]) {
			ret = -ENOMEM;
			mlog_errno(ret);
			goto out;
		}
	}

	if (new) {
		/*
		 * This is safe to call with the page locks - it won't take
		 * any additional semaphores or cluster locks.
		 */
		tmp_pos = wc->w_cpos;
		ret = ocfs2_do_extend_allocation(OCFS2_SB(inode->i_sb), inode,
						 &tmp_pos, 1, di_bh, handle,
						 data_ac, meta_ac, NULL);
		/*
		 * This shouldn't happen because we must have already
		 * calculated the correct meta data allocation required. The
		 * internal tree allocation code should know how to increase
		 * transaction credits itself.
		 *
		 * If need be, we could handle -EAGAIN for a
		 * RESTART_TRANS here.
		 */
		mlog_bug_on_msg(ret == -EAGAIN,
				"Inode %llu: EAGAIN return during allocation.\n",
				(unsigned long long)OCFS2_I(inode)->ip_blkno);
		if (ret < 0) {
			mlog_errno(ret);
			goto out;
		}
	}

	ret = ocfs2_extent_map_get_blocks(inode, v_blkno, &p_blkno, NULL);
	if (ret < 0) {

		/*
		 * XXX: Should we go readonly here?
		 */

		mlog_errno(ret);
		goto out;
	}

	BUG_ON(p_blkno == 0);

	for(i = 0; i < numpages; i++) {
		ret = ocfs2_write_data_page(inode, handle, &p_blkno, cpages[i],
					    wc, new);
		if (ret < 0) {
			mlog_errno(ret);
			goto out;
		}

		copied += ret;
	}

out:
	for(i = 0; i < numpages; i++) {
		unlock_page(cpages[i]);
		mark_page_accessed(cpages[i]);
		page_cache_release(cpages[i]);
	}
	kfree(cpages);

	return copied ? copied : ret;
}

static void ocfs2_write_ctxt_init(struct ocfs2_write_ctxt *wc,
				  struct ocfs2_super *osb, loff_t pos,
				  size_t count, ocfs2_page_writer *cb,
				  void *cb_priv)
{
	wc->w_count = count;
	wc->w_pos = pos;
	wc->w_cpos = wc->w_pos >> osb->s_clustersize_bits;
	wc->w_finished_copy = 0;

	if (unlikely(PAGE_CACHE_SHIFT > osb->s_clustersize_bits))
		wc->w_large_pages = 1;
	else
		wc->w_large_pages = 0;

	wc->w_write_data_page = cb;
	wc->w_private = cb_priv;
}

/*
 * Write a cluster to an inode. The cluster may not be allocated yet,
 * in which case it will be. This only exists for buffered writes -
 * O_DIRECT takes a more "traditional" path through the kernel.
 *
 * The caller is responsible for incrementing pos, written counts, etc
 *
 * For file systems that don't support sparse files, pre-allocation
 * and page zeroing up until cpos should be done prior to this
 * function call.
 *
 * Callers should be holding i_sem, and the rw cluster lock.
 *
 * Returns the number of user bytes written, or less than zero for
 * error.
 */
ssize_t ocfs2_buffered_write_cluster(struct file *file, loff_t pos,
				     size_t count, ocfs2_page_writer *actor,
				     void *priv)
{
	int ret, credits = OCFS2_INODE_UPDATE_CREDITS;
	ssize_t written = 0;
	u32 phys;
	struct inode *inode = file->f_mapping->host;
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	struct buffer_head *di_bh = NULL;
	struct ocfs2_dinode *di;
	struct ocfs2_alloc_context *data_ac = NULL;
	struct ocfs2_alloc_context *meta_ac = NULL;
	handle_t *handle;
	struct ocfs2_write_ctxt wc;

	ocfs2_write_ctxt_init(&wc, osb, pos, count, actor, priv);

	ret = ocfs2_meta_lock(inode, &di_bh, 1);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}
	di = (struct ocfs2_dinode *)di_bh->b_data;

	/*
	 * Take alloc sem here to prevent concurrent lookups. That way
	 * the mapping, zeroing and tree manipulation within
	 * ocfs2_write() will be safe against ->readpage(). This
	 * should also serve to lock out allocation from a shared
	 * writeable region.
	 */
	down_write(&OCFS2_I(inode)->ip_alloc_sem);

	ret = ocfs2_get_clusters(inode, wc.w_cpos, &phys, NULL);
	if (ret) {
		mlog_errno(ret);
		goto out_meta;
	}

	/* phys == 0 means that allocation is required. */
	if (phys == 0) {
		ret = ocfs2_lock_allocators(inode, di, 1, &data_ac, &meta_ac);
		if (ret) {
			mlog_errno(ret);
			goto out_meta;
		}

		credits = ocfs2_calc_extend_credits(inode->i_sb, di, 1);
	}

	ret = ocfs2_data_lock(inode, 1);
	if (ret) {
		mlog_errno(ret);
		goto out_meta;
	}

	handle = ocfs2_start_trans(osb, credits);
	if (IS_ERR(handle)) {
		ret = PTR_ERR(handle);
		mlog_errno(ret);
		goto out_data;
	}

	written = ocfs2_write(file, phys, handle, di_bh, data_ac,
			      meta_ac, &wc);
	if (written < 0) {
		ret = written;
		mlog_errno(ret);
		goto out_commit;
	}

	ret = ocfs2_journal_access(handle, inode, di_bh,
				   OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret) {
		mlog_errno(ret);
		goto out_commit;
	}

	pos += written;
	if (pos > inode->i_size) {
		i_size_write(inode, pos);
		mark_inode_dirty(inode);
	}
	inode->i_blocks = ocfs2_align_bytes_to_sectors((u64)(i_size_read(inode)));
	di->i_size = cpu_to_le64((u64)i_size_read(inode));
	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
	di->i_mtime = di->i_ctime = cpu_to_le64(inode->i_mtime.tv_sec);
	di->i_mtime_nsec = di->i_ctime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);

	ret = ocfs2_journal_dirty(handle, di_bh);
	if (ret)
		mlog_errno(ret);

out_commit:
	ocfs2_commit_trans(osb, handle);

out_data:
	ocfs2_data_unlock(inode, 1);

out_meta:
	up_write(&OCFS2_I(inode)->ip_alloc_sem);
	ocfs2_meta_unlock(inode, 1);

out:
	brelse(di_bh);
	if (data_ac)
		ocfs2_free_alloc_context(data_ac);
	if (meta_ac)
		ocfs2_free_alloc_context(meta_ac);

	return written ? written : ret;
}

const struct address_space_operations ocfs2_aops = {
	.readpage	= ocfs2_readpage,
	.writepage	= ocfs2_writepage,
	.bmap		= ocfs2_bmap,
	.sync_page	= block_sync_page,
	.direct_IO	= ocfs2_direct_IO,
	.invalidatepage	= ocfs2_invalidatepage,
	.releasepage	= ocfs2_releasepage,
	.migratepage	= buffer_migrate_page,
};