alloc.c 96.2 KB
Newer Older
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
/* -*- mode: c; c-basic-offset: 8; -*-
 * vim: noexpandtab sw=8 ts=8 sts=0:
 *
 * alloc.c
 *
 * Extent allocs and frees
 *
 * 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/types.h>
#include <linux/slab.h>
#include <linux/highmem.h>
30
#include <linux/swap.h>
31
32
33
34
35
36
37

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

#include "ocfs2.h"

#include "alloc.h"
38
#include "aops.h"
39
40
41
42
43
44
45
46
47
48
49
50
51
#include "dlmglue.h"
#include "extent_map.h"
#include "inode.h"
#include "journal.h"
#include "localalloc.h"
#include "suballoc.h"
#include "sysfile.h"
#include "file.h"
#include "super.h"
#include "uptodate.h"

#include "buffer_head_io.h"

Mark Fasheh's avatar
Mark Fasheh committed
52
static void ocfs2_free_truncate_context(struct ocfs2_truncate_context *tc);
53

Mark Fasheh's avatar
Mark Fasheh committed
54
55
56
57
58
59
60
61
62
63
64
/*
 * Structures which describe a path through a btree, and functions to
 * manipulate them.
 *
 * The idea here is to be as generic as possible with the tree
 * manipulation code.
 */
struct ocfs2_path_item {
	struct buffer_head		*bh;
	struct ocfs2_extent_list	*el;
};
65

Mark Fasheh's avatar
Mark Fasheh committed
66
#define OCFS2_MAX_PATH_DEPTH	5
67

Mark Fasheh's avatar
Mark Fasheh committed
68
69
70
71
struct ocfs2_path {
	int			p_tree_depth;
	struct ocfs2_path_item	p_node[OCFS2_MAX_PATH_DEPTH];
};
72

Mark Fasheh's avatar
Mark Fasheh committed
73
74
75
76
77
#define path_root_bh(_path) ((_path)->p_node[0].bh)
#define path_root_el(_path) ((_path)->p_node[0].el)
#define path_leaf_bh(_path) ((_path)->p_node[(_path)->p_tree_depth].bh)
#define path_leaf_el(_path) ((_path)->p_node[(_path)->p_tree_depth].el)
#define path_num_items(_path) ((_path)->p_tree_depth + 1)
78

Mark Fasheh's avatar
Mark Fasheh committed
79
80
81
82
83
84
85
86
87
/*
 * Reset the actual path elements so that we can re-use the structure
 * to build another path. Generally, this involves freeing the buffer
 * heads.
 */
static void ocfs2_reinit_path(struct ocfs2_path *path, int keep_root)
{
	int i, start = 0, depth = 0;
	struct ocfs2_path_item *node;
88

Mark Fasheh's avatar
Mark Fasheh committed
89
90
	if (keep_root)
		start = 1;
91

Mark Fasheh's avatar
Mark Fasheh committed
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
	for(i = start; i < path_num_items(path); i++) {
		node = &path->p_node[i];

		brelse(node->bh);
		node->bh = NULL;
		node->el = NULL;
	}

	/*
	 * Tree depth may change during truncate, or insert. If we're
	 * keeping the root extent list, then make sure that our path
	 * structure reflects the proper depth.
	 */
	if (keep_root)
		depth = le16_to_cpu(path_root_el(path)->l_tree_depth);

	path->p_tree_depth = depth;
}

static void ocfs2_free_path(struct ocfs2_path *path)
{
	if (path) {
		ocfs2_reinit_path(path, 0);
		kfree(path);
	}
}

/*
 * Make the *dest path the same as src and re-initialize src path to
 * have a root only.
 */
static void ocfs2_mv_path(struct ocfs2_path *dest, struct ocfs2_path *src)
{
	int i;

	BUG_ON(path_root_bh(dest) != path_root_bh(src));

	for(i = 1; i < OCFS2_MAX_PATH_DEPTH; i++) {
		brelse(dest->p_node[i].bh);

		dest->p_node[i].bh = src->p_node[i].bh;
		dest->p_node[i].el = src->p_node[i].el;

		src->p_node[i].bh = NULL;
		src->p_node[i].el = NULL;
	}
}

/*
 * Insert an extent block at given index.
 *
 * This will not take an additional reference on eb_bh.
 */
static inline void ocfs2_path_insert_eb(struct ocfs2_path *path, int index,
					struct buffer_head *eb_bh)
{
	struct ocfs2_extent_block *eb = (struct ocfs2_extent_block *)eb_bh->b_data;

	/*
	 * Right now, no root bh is an extent block, so this helps
	 * catch code errors with dinode trees. The assertion can be
	 * safely removed if we ever need to insert extent block
	 * structures at the root.
	 */
	BUG_ON(index == 0);

	path->p_node[index].bh = eb_bh;
	path->p_node[index].el = &eb->h_list;
}

static struct ocfs2_path *ocfs2_new_path(struct buffer_head *root_bh,
					 struct ocfs2_extent_list *root_el)
{
	struct ocfs2_path *path;
166

Mark Fasheh's avatar
Mark Fasheh committed
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
	BUG_ON(le16_to_cpu(root_el->l_tree_depth) >= OCFS2_MAX_PATH_DEPTH);

	path = kzalloc(sizeof(*path), GFP_NOFS);
	if (path) {
		path->p_tree_depth = le16_to_cpu(root_el->l_tree_depth);
		get_bh(root_bh);
		path_root_bh(path) = root_bh;
		path_root_el(path) = root_el;
	}

	return path;
}

/*
 * Allocate and initialize a new path based on a disk inode tree.
 */
static struct ocfs2_path *ocfs2_new_inode_path(struct buffer_head *di_bh)
{
	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
	struct ocfs2_extent_list *el = &di->id2.i_list;

	return ocfs2_new_path(di_bh, el);
}

/*
 * Convenience function to journal all components in a path.
 */
static int ocfs2_journal_access_path(struct inode *inode, handle_t *handle,
				     struct ocfs2_path *path)
{
	int i, ret = 0;

	if (!path)
		goto out;

	for(i = 0; i < path_num_items(path); i++) {
		ret = ocfs2_journal_access(handle, inode, path->p_node[i].bh,
					   OCFS2_JOURNAL_ACCESS_WRITE);
		if (ret < 0) {
			mlog_errno(ret);
			goto out;
		}
	}

out:
	return ret;
}

enum ocfs2_contig_type {
	CONTIG_NONE = 0,
	CONTIG_LEFT,
	CONTIG_RIGHT
};

static int ocfs2_block_extent_contig(struct super_block *sb,
				     struct ocfs2_extent_rec *ext,
				     u64 blkno)
224
225
{
	return blkno == (le64_to_cpu(ext->e_blkno) +
Mark Fasheh's avatar
Mark Fasheh committed
226
			 ocfs2_clusters_to_blocks(sb,
227
228
229
						  le32_to_cpu(ext->e_clusters)));
}

Mark Fasheh's avatar
Mark Fasheh committed
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
static int ocfs2_extents_adjacent(struct ocfs2_extent_rec *left,
				  struct ocfs2_extent_rec *right)
{
	return (le32_to_cpu(left->e_cpos) + le32_to_cpu(left->e_clusters) ==
		le32_to_cpu(right->e_cpos));
}

static enum ocfs2_contig_type
	ocfs2_extent_contig(struct inode *inode,
			    struct ocfs2_extent_rec *ext,
			    struct ocfs2_extent_rec *insert_rec)
{
	u64 blkno = le64_to_cpu(insert_rec->e_blkno);

	if (ocfs2_extents_adjacent(ext, insert_rec) &&
	    ocfs2_block_extent_contig(inode->i_sb, ext, blkno))
			return CONTIG_RIGHT;

	blkno = le64_to_cpu(ext->e_blkno);
	if (ocfs2_extents_adjacent(insert_rec, ext) &&
	    ocfs2_block_extent_contig(inode->i_sb, insert_rec, blkno))
		return CONTIG_LEFT;

	return CONTIG_NONE;
}

/*
 * NOTE: We can have pretty much any combination of contiguousness and
 * appending.
 *
 * The usefulness of APPEND_TAIL is more in that it lets us know that
 * we'll have to update the path to that leaf.
 */
enum ocfs2_append_type {
	APPEND_NONE = 0,
	APPEND_TAIL,
};

struct ocfs2_insert_type {
	enum ocfs2_append_type	ins_appending;
	enum ocfs2_contig_type	ins_contig;
	int			ins_contig_index;
	int			ins_free_records;
	int			ins_tree_depth;
};

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
/*
 * How many free extents have we got before we need more meta data?
 */
int ocfs2_num_free_extents(struct ocfs2_super *osb,
			   struct inode *inode,
			   struct ocfs2_dinode *fe)
{
	int retval;
	struct ocfs2_extent_list *el;
	struct ocfs2_extent_block *eb;
	struct buffer_head *eb_bh = NULL;

	mlog_entry_void();

	if (!OCFS2_IS_VALID_DINODE(fe)) {
		OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, fe);
		retval = -EIO;
		goto bail;
	}

	if (fe->i_last_eb_blk) {
		retval = ocfs2_read_block(osb, le64_to_cpu(fe->i_last_eb_blk),
					  &eb_bh, OCFS2_BH_CACHED, inode);
		if (retval < 0) {
			mlog_errno(retval);
			goto bail;
		}
		eb = (struct ocfs2_extent_block *) eb_bh->b_data;
		el = &eb->h_list;
	} else
		el = &fe->id2.i_list;

	BUG_ON(el->l_tree_depth != 0);

	retval = le16_to_cpu(el->l_count) - le16_to_cpu(el->l_next_free_rec);
bail:
	if (eb_bh)
		brelse(eb_bh);

	mlog_exit(retval);
	return retval;
}

/* expects array to already be allocated
 *
 * sets h_signature, h_blkno, h_suballoc_bit, h_suballoc_slot, and
 * l_count for you
 */
static int ocfs2_create_new_meta_bhs(struct ocfs2_super *osb,
325
				     handle_t *handle,
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
				     struct inode *inode,
				     int wanted,
				     struct ocfs2_alloc_context *meta_ac,
				     struct buffer_head *bhs[])
{
	int count, status, i;
	u16 suballoc_bit_start;
	u32 num_got;
	u64 first_blkno;
	struct ocfs2_extent_block *eb;

	mlog_entry_void();

	count = 0;
	while (count < wanted) {
		status = ocfs2_claim_metadata(osb,
					      handle,
					      meta_ac,
					      wanted - count,
					      &suballoc_bit_start,
					      &num_got,
					      &first_blkno);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}

		for(i = count;  i < (num_got + count); i++) {
			bhs[i] = sb_getblk(osb->sb, first_blkno);
			if (bhs[i] == NULL) {
				status = -EIO;
				mlog_errno(status);
				goto bail;
			}
			ocfs2_set_new_buffer_uptodate(inode, bhs[i]);

			status = ocfs2_journal_access(handle, inode, bhs[i],
						      OCFS2_JOURNAL_ACCESS_CREATE);
			if (status < 0) {
				mlog_errno(status);
				goto bail;
			}

			memset(bhs[i]->b_data, 0, osb->sb->s_blocksize);
			eb = (struct ocfs2_extent_block *) bhs[i]->b_data;
			/* Ok, setup the minimal stuff here. */
			strcpy(eb->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE);
			eb->h_blkno = cpu_to_le64(first_blkno);
			eb->h_fs_generation = cpu_to_le32(osb->fs_generation);

#ifndef OCFS2_USE_ALL_METADATA_SUBALLOCATORS
			/* we always use slot zero's suballocator */
			eb->h_suballoc_slot = 0;
#else
			eb->h_suballoc_slot = cpu_to_le16(osb->slot_num);
#endif
			eb->h_suballoc_bit = cpu_to_le16(suballoc_bit_start);
			eb->h_list.l_count =
				cpu_to_le16(ocfs2_extent_recs_per_eb(osb->sb));

			suballoc_bit_start++;
			first_blkno++;

			/* We'll also be dirtied by the caller, so
			 * this isn't absolutely necessary. */
			status = ocfs2_journal_dirty(handle, bhs[i]);
			if (status < 0) {
				mlog_errno(status);
				goto bail;
			}
		}

		count += num_got;
	}

	status = 0;
bail:
	if (status < 0) {
		for(i = 0; i < wanted; i++) {
			if (bhs[i])
				brelse(bhs[i]);
			bhs[i] = NULL;
		}
	}
	mlog_exit(status);
	return status;
}

Mark Fasheh's avatar
Mark Fasheh committed
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
/*
 * Helper function for ocfs2_add_branch() and ocfs2_shift_tree_depth().
 *
 * Returns the sum of the rightmost extent rec logical offset and
 * cluster count.
 *
 * ocfs2_add_branch() uses this to determine what logical cluster
 * value should be populated into the leftmost new branch records.
 *
 * ocfs2_shift_tree_depth() uses this to determine the # clusters
 * value for the new topmost tree record.
 */
static inline u32 ocfs2_sum_rightmost_rec(struct ocfs2_extent_list  *el)
{
	int i;

	i = le16_to_cpu(el->l_next_free_rec) - 1;

	return le32_to_cpu(el->l_recs[i].e_cpos) +
		le32_to_cpu(el->l_recs[i].e_clusters);
}

436
437
438
439
440
441
442
443
444
445
446
447
/*
 * Add an entire tree branch to our inode. eb_bh is the extent block
 * to start at, if we don't want to start the branch at the dinode
 * structure.
 *
 * last_eb_bh is required as we have to update it's next_leaf pointer
 * for the new last extent block.
 *
 * the new branch will be 'empty' in the sense that every block will
 * contain a single record with e_clusters == 0.
 */
static int ocfs2_add_branch(struct ocfs2_super *osb,
448
			    handle_t *handle,
449
450
451
452
453
454
455
456
457
458
459
460
461
462
			    struct inode *inode,
			    struct buffer_head *fe_bh,
			    struct buffer_head *eb_bh,
			    struct buffer_head *last_eb_bh,
			    struct ocfs2_alloc_context *meta_ac)
{
	int status, new_blocks, i;
	u64 next_blkno, new_last_eb_blk;
	struct buffer_head *bh;
	struct buffer_head **new_eb_bhs = NULL;
	struct ocfs2_dinode *fe;
	struct ocfs2_extent_block *eb;
	struct ocfs2_extent_list  *eb_el;
	struct ocfs2_extent_list  *el;
Mark Fasheh's avatar
Mark Fasheh committed
463
	u32 new_cpos;
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

	mlog_entry_void();

	BUG_ON(!last_eb_bh);

	fe = (struct ocfs2_dinode *) fe_bh->b_data;

	if (eb_bh) {
		eb = (struct ocfs2_extent_block *) eb_bh->b_data;
		el = &eb->h_list;
	} else
		el = &fe->id2.i_list;

	/* we never add a branch to a leaf. */
	BUG_ON(!el->l_tree_depth);

	new_blocks = le16_to_cpu(el->l_tree_depth);

	/* allocate the number of new eb blocks we need */
	new_eb_bhs = kcalloc(new_blocks, sizeof(struct buffer_head *),
			     GFP_KERNEL);
	if (!new_eb_bhs) {
		status = -ENOMEM;
		mlog_errno(status);
		goto bail;
	}

	status = ocfs2_create_new_meta_bhs(osb, handle, inode, new_blocks,
					   meta_ac, new_eb_bhs);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

Mark Fasheh's avatar
Mark Fasheh committed
498
499
500
	eb = (struct ocfs2_extent_block *)last_eb_bh->b_data;
	new_cpos = ocfs2_sum_rightmost_rec(&eb->h_list);

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
	/* Note: new_eb_bhs[new_blocks - 1] is the guy which will be
	 * linked with the rest of the tree.
	 * conversly, new_eb_bhs[0] is the new bottommost leaf.
	 *
	 * when we leave the loop, new_last_eb_blk will point to the
	 * newest leaf, and next_blkno will point to the topmost extent
	 * block. */
	next_blkno = new_last_eb_blk = 0;
	for(i = 0; i < new_blocks; i++) {
		bh = new_eb_bhs[i];
		eb = (struct ocfs2_extent_block *) bh->b_data;
		if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
			OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
			status = -EIO;
			goto bail;
		}
		eb_el = &eb->h_list;

		status = ocfs2_journal_access(handle, inode, bh,
					      OCFS2_JOURNAL_ACCESS_CREATE);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}

		eb->h_next_leaf_blk = 0;
		eb_el->l_tree_depth = cpu_to_le16(i);
		eb_el->l_next_free_rec = cpu_to_le16(1);
Mark Fasheh's avatar
Mark Fasheh committed
529
530
531
532
533
		/*
		 * This actually counts as an empty extent as
		 * c_clusters == 0
		 */
		eb_el->l_recs[0].e_cpos = cpu_to_le32(new_cpos);
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
		eb_el->l_recs[0].e_blkno = cpu_to_le64(next_blkno);
		eb_el->l_recs[0].e_clusters = cpu_to_le32(0);
		if (!eb_el->l_tree_depth)
			new_last_eb_blk = le64_to_cpu(eb->h_blkno);

		status = ocfs2_journal_dirty(handle, bh);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}

		next_blkno = le64_to_cpu(eb->h_blkno);
	}

	/* This is a bit hairy. We want to update up to three blocks
	 * here without leaving any of them in an inconsistent state
	 * in case of error. We don't have to worry about
	 * journal_dirty erroring as it won't unless we've aborted the
	 * handle (in which case we would never be here) so reserving
	 * the write with journal_access is all we need to do. */
	status = ocfs2_journal_access(handle, inode, last_eb_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}
	status = ocfs2_journal_access(handle, inode, fe_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}
	if (eb_bh) {
		status = ocfs2_journal_access(handle, inode, eb_bh,
					      OCFS2_JOURNAL_ACCESS_WRITE);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}
	}

	/* Link the new branch into the rest of the tree (el will
	 * either be on the fe, or the extent block passed in. */
	i = le16_to_cpu(el->l_next_free_rec);
	el->l_recs[i].e_blkno = cpu_to_le64(next_blkno);
Mark Fasheh's avatar
Mark Fasheh committed
579
	el->l_recs[i].e_cpos = cpu_to_le32(new_cpos);
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
	el->l_recs[i].e_clusters = 0;
	le16_add_cpu(&el->l_next_free_rec, 1);

	/* fe needs a new last extent block pointer, as does the
	 * next_leaf on the previously last-extent-block. */
	fe->i_last_eb_blk = cpu_to_le64(new_last_eb_blk);

	eb = (struct ocfs2_extent_block *) last_eb_bh->b_data;
	eb->h_next_leaf_blk = cpu_to_le64(new_last_eb_blk);

	status = ocfs2_journal_dirty(handle, last_eb_bh);
	if (status < 0)
		mlog_errno(status);
	status = ocfs2_journal_dirty(handle, fe_bh);
	if (status < 0)
		mlog_errno(status);
	if (eb_bh) {
		status = ocfs2_journal_dirty(handle, eb_bh);
		if (status < 0)
			mlog_errno(status);
	}

	status = 0;
bail:
	if (new_eb_bhs) {
		for (i = 0; i < new_blocks; i++)
			if (new_eb_bhs[i])
				brelse(new_eb_bhs[i]);
		kfree(new_eb_bhs);
	}

	mlog_exit(status);
	return status;
}

/*
 * adds another level to the allocation tree.
 * returns back the new extent block so you can add a branch to it
 * after this call.
 */
static int ocfs2_shift_tree_depth(struct ocfs2_super *osb,
621
				  handle_t *handle,
622
623
624
625
626
627
				  struct inode *inode,
				  struct buffer_head *fe_bh,
				  struct ocfs2_alloc_context *meta_ac,
				  struct buffer_head **ret_new_eb_bh)
{
	int status, i;
Mark Fasheh's avatar
Mark Fasheh committed
628
	u32 new_clusters;
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
	struct buffer_head *new_eb_bh = NULL;
	struct ocfs2_dinode *fe;
	struct ocfs2_extent_block *eb;
	struct ocfs2_extent_list  *fe_el;
	struct ocfs2_extent_list  *eb_el;

	mlog_entry_void();

	status = ocfs2_create_new_meta_bhs(osb, handle, inode, 1, meta_ac,
					   &new_eb_bh);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	eb = (struct ocfs2_extent_block *) new_eb_bh->b_data;
	if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
		OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
		status = -EIO;
		goto bail;
	}

	eb_el = &eb->h_list;
	fe = (struct ocfs2_dinode *) fe_bh->b_data;
	fe_el = &fe->id2.i_list;

	status = ocfs2_journal_access(handle, inode, new_eb_bh,
				      OCFS2_JOURNAL_ACCESS_CREATE);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	/* copy the fe data into the new extent block */
	eb_el->l_tree_depth = fe_el->l_tree_depth;
	eb_el->l_next_free_rec = fe_el->l_next_free_rec;
	for(i = 0; i < le16_to_cpu(fe_el->l_next_free_rec); i++) {
		eb_el->l_recs[i].e_cpos = fe_el->l_recs[i].e_cpos;
		eb_el->l_recs[i].e_clusters = fe_el->l_recs[i].e_clusters;
		eb_el->l_recs[i].e_blkno = fe_el->l_recs[i].e_blkno;
	}

	status = ocfs2_journal_dirty(handle, new_eb_bh);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	status = ocfs2_journal_access(handle, inode, fe_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

Mark Fasheh's avatar
Mark Fasheh committed
684
685
	new_clusters = ocfs2_sum_rightmost_rec(eb_el);

686
687
688
689
	/* update fe now */
	le16_add_cpu(&fe_el->l_tree_depth, 1);
	fe_el->l_recs[0].e_cpos = 0;
	fe_el->l_recs[0].e_blkno = eb->h_blkno;
Mark Fasheh's avatar
Mark Fasheh committed
690
	fe_el->l_recs[0].e_clusters = cpu_to_le32(new_clusters);
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
	for(i = 1; i < le16_to_cpu(fe_el->l_next_free_rec); i++) {
		fe_el->l_recs[i].e_cpos = 0;
		fe_el->l_recs[i].e_clusters = 0;
		fe_el->l_recs[i].e_blkno = 0;
	}
	fe_el->l_next_free_rec = cpu_to_le16(1);

	/* If this is our 1st tree depth shift, then last_eb_blk
	 * becomes the allocated extent block */
	if (fe_el->l_tree_depth == cpu_to_le16(1))
		fe->i_last_eb_blk = eb->h_blkno;

	status = ocfs2_journal_dirty(handle, fe_bh);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	*ret_new_eb_bh = new_eb_bh;
	new_eb_bh = NULL;
	status = 0;
bail:
	if (new_eb_bh)
		brelse(new_eb_bh);

	mlog_exit(status);
	return status;
}

/*
 * Should only be called when there is no space left in any of the
 * leaf nodes. What we want to do is find the lowest tree depth
 * non-leaf extent block with room for new records. There are three
 * valid results of this search:
 *
 * 1) a lowest extent block is found, then we pass it back in
 *    *lowest_eb_bh and return '0'
 *
 * 2) the search fails to find anything, but the dinode has room. We
 *    pass NULL back in *lowest_eb_bh, but still return '0'
 *
 * 3) the search fails to find anything AND the dinode is full, in
 *    which case we return > 0
 *
 * return status < 0 indicates an error.
 */
static int ocfs2_find_branch_target(struct ocfs2_super *osb,
				    struct inode *inode,
				    struct buffer_head *fe_bh,
				    struct buffer_head **target_bh)
{
	int status = 0, i;
	u64 blkno;
	struct ocfs2_dinode *fe;
	struct ocfs2_extent_block *eb;
	struct ocfs2_extent_list  *el;
	struct buffer_head *bh = NULL;
	struct buffer_head *lowest_bh = NULL;

	mlog_entry_void();

	*target_bh = NULL;

	fe = (struct ocfs2_dinode *) fe_bh->b_data;
	el = &fe->id2.i_list;

	while(le16_to_cpu(el->l_tree_depth) > 1) {
		if (le16_to_cpu(el->l_next_free_rec) == 0) {
759
			ocfs2_error(inode->i_sb, "Dinode %llu has empty "
760
				    "extent list (next_free_rec == 0)",
761
				    (unsigned long long)OCFS2_I(inode)->ip_blkno);
762
763
764
765
766
767
			status = -EIO;
			goto bail;
		}
		i = le16_to_cpu(el->l_next_free_rec) - 1;
		blkno = le64_to_cpu(el->l_recs[i].e_blkno);
		if (!blkno) {
768
			ocfs2_error(inode->i_sb, "Dinode %llu has extent "
769
770
				    "list where extent # %d has no physical "
				    "block start",
771
				    (unsigned long long)OCFS2_I(inode)->ip_blkno, i);
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
			status = -EIO;
			goto bail;
		}

		if (bh) {
			brelse(bh);
			bh = NULL;
		}

		status = ocfs2_read_block(osb, blkno, &bh, OCFS2_BH_CACHED,
					  inode);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}
Mark Fasheh's avatar
Mark Fasheh committed
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

		eb = (struct ocfs2_extent_block *) bh->b_data;
		if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
			OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
			status = -EIO;
			goto bail;
		}
		el = &eb->h_list;

		if (le16_to_cpu(el->l_next_free_rec) <
		    le16_to_cpu(el->l_count)) {
			if (lowest_bh)
				brelse(lowest_bh);
			lowest_bh = bh;
			get_bh(lowest_bh);
		}
	}

	/* If we didn't find one and the fe doesn't have any room,
	 * then return '1' */
	if (!lowest_bh
	    && (fe->id2.i_list.l_next_free_rec == fe->id2.i_list.l_count))
		status = 1;

	*target_bh = lowest_bh;
bail:
	if (bh)
		brelse(bh);

	mlog_exit(status);
	return status;
}

static inline int ocfs2_is_empty_extent(struct ocfs2_extent_rec *rec)
{
	return !rec->e_clusters;
}

/*
 * This function will discard the rightmost extent record.
 */
static void ocfs2_shift_records_right(struct ocfs2_extent_list *el)
{
	int next_free = le16_to_cpu(el->l_next_free_rec);
	int count = le16_to_cpu(el->l_count);
	unsigned int num_bytes;

	BUG_ON(!next_free);
	/* This will cause us to go off the end of our extent list. */
	BUG_ON(next_free >= count);

	num_bytes = sizeof(struct ocfs2_extent_rec) * next_free;

	memmove(&el->l_recs[1], &el->l_recs[0], num_bytes);
}

static void ocfs2_rotate_leaf(struct ocfs2_extent_list *el,
			      struct ocfs2_extent_rec *insert_rec)
{
	int i, insert_index, next_free, has_empty, num_bytes;
	u32 insert_cpos = le32_to_cpu(insert_rec->e_cpos);
	struct ocfs2_extent_rec *rec;

	next_free = le16_to_cpu(el->l_next_free_rec);
	has_empty = ocfs2_is_empty_extent(&el->l_recs[0]);

	BUG_ON(!next_free);

	/* The tree code before us didn't allow enough room in the leaf. */
	if (el->l_next_free_rec == el->l_count && !has_empty)
		BUG();

	/*
	 * The easiest way to approach this is to just remove the
	 * empty extent and temporarily decrement next_free.
	 */
	if (has_empty) {
		/*
		 * If next_free was 1 (only an empty extent), this
		 * loop won't execute, which is fine. We still want
		 * the decrement above to happen.
		 */
		for(i = 0; i < (next_free - 1); i++)
			el->l_recs[i] = el->l_recs[i+1];

		next_free--;
	}

	/*
	 * Figure out what the new record index should be.
	 */
	for(i = 0; i < next_free; i++) {
		rec = &el->l_recs[i];

		if (insert_cpos < le32_to_cpu(rec->e_cpos))
			break;
	}
	insert_index = i;

	mlog(0, "ins %u: index %d, has_empty %d, next_free %d, count %d\n",
	     insert_cpos, insert_index, has_empty, next_free, le16_to_cpu(el->l_count));

	BUG_ON(insert_index < 0);
	BUG_ON(insert_index >= le16_to_cpu(el->l_count));
	BUG_ON(insert_index > next_free);

	/*
	 * No need to memmove if we're just adding to the tail.
	 */
	if (insert_index != next_free) {
		BUG_ON(next_free >= le16_to_cpu(el->l_count));

		num_bytes = next_free - insert_index;
		num_bytes *= sizeof(struct ocfs2_extent_rec);
		memmove(&el->l_recs[insert_index + 1],
			&el->l_recs[insert_index],
			num_bytes);
	}

	/*
	 * Either we had an empty extent, and need to re-increment or
	 * there was no empty extent on a non full rightmost leaf node,
	 * in which case we still need to increment.
	 */
	next_free++;
	el->l_next_free_rec = cpu_to_le16(next_free);
	/*
	 * Make sure none of the math above just messed up our tree.
	 */
	BUG_ON(le16_to_cpu(el->l_next_free_rec) > le16_to_cpu(el->l_count));

	el->l_recs[insert_index] = *insert_rec;

}

/*
 * Create an empty extent record .
 *
 * l_next_free_rec may be updated.
 *
 * If an empty extent already exists do nothing.
 */
static void ocfs2_create_empty_extent(struct ocfs2_extent_list *el)
{
	int next_free = le16_to_cpu(el->l_next_free_rec);

	if (next_free == 0)
		goto set_and_inc;

	if (ocfs2_is_empty_extent(&el->l_recs[0]))
		return;

	mlog_bug_on_msg(el->l_count == el->l_next_free_rec,
			"Asked to create an empty extent in a full list:\n"
			"count = %u, tree depth = %u",
			le16_to_cpu(el->l_count),
			le16_to_cpu(el->l_tree_depth));

	ocfs2_shift_records_right(el);

set_and_inc:
	le16_add_cpu(&el->l_next_free_rec, 1);
	memset(&el->l_recs[0], 0, sizeof(struct ocfs2_extent_rec));
}

/*
 * For a rotation which involves two leaf nodes, the "root node" is
 * the lowest level tree node which contains a path to both leafs. This
 * resulting set of information can be used to form a complete "subtree"
 *
 * This function is passed two full paths from the dinode down to a
 * pair of adjacent leaves. It's task is to figure out which path
 * index contains the subtree root - this can be the root index itself
 * in a worst-case rotation.
 *
 * The array index of the subtree root is passed back.
 */
static int ocfs2_find_subtree_root(struct inode *inode,
				   struct ocfs2_path *left,
				   struct ocfs2_path *right)
{
	int i = 0;

	/*
	 * Check that the caller passed in two paths from the same tree.
	 */
	BUG_ON(path_root_bh(left) != path_root_bh(right));

	do {
		i++;

		/*
		 * The caller didn't pass two adjacent paths.
		 */
		mlog_bug_on_msg(i > left->p_tree_depth,
				"Inode %lu, left depth %u, right depth %u\n"
				"left leaf blk %llu, right leaf blk %llu\n",
				inode->i_ino, left->p_tree_depth,
				right->p_tree_depth,
				(unsigned long long)path_leaf_bh(left)->b_blocknr,
				(unsigned long long)path_leaf_bh(right)->b_blocknr);
	} while (left->p_node[i].bh->b_blocknr ==
		 right->p_node[i].bh->b_blocknr);

	return i - 1;
}

typedef void (path_insert_t)(void *, struct buffer_head *);

/*
 * Traverse a btree path in search of cpos, starting at root_el.
 *
 * This code can be called with a cpos larger than the tree, in which
 * case it will return the rightmost path.