aboutsummaryrefslogtreecommitdiffstats
path: root/fs/btrfs/inode-map.c
blob: ee15d88b33d2a3dca65f41a4c5c3f6a36bbe5b48 (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
/*
 * Copyright (C) 2007 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 v2 as published by the Free Software Foundation.
 *
 * 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/delay.h>
#include <linux/kthread.h>
#include <linux/pagemap.h>

#include "ctree.h"
#include "disk-io.h"
#include "free-space-cache.h"
#include "inode-map.h"
#include "transaction.h"

static int caching_kthread(void *data)
{
	struct btrfs_root *root = data;
	struct btrfs_fs_info *fs_info = root->fs_info;
	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
	struct btrfs_key key;
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	u64 last = (u64)-1;
	int slot;
	int ret;

	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
		return 0;

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;

	/* Since the commit root is read-only, we can safely skip locking. */
	path->skip_locking = 1;
	path->search_commit_root = 1;
	path->reada = 2;

	key.objectid = BTRFS_FIRST_FREE_OBJECTID;
	key.offset = 0;
	key.type = BTRFS_INODE_ITEM_KEY;
again:
	/* need to make sure the commit_root doesn't disappear */
	mutex_lock(&root->fs_commit_mutex);

	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
	if (ret < 0)
		goto out;

	while (1) {
		if (btrfs_fs_closing(fs_info))
			goto out;

		leaf = path->nodes[0];
		slot = path->slots[0];
		if (slot >= btrfs_header_nritems(leaf)) {
			ret = btrfs_next_leaf(root, path);
			if (ret < 0)
				goto out;
			else if (ret > 0)
				break;

			if (need_resched() ||
			    btrfs_transaction_in_commit(fs_info)) {
				leaf = path->nodes[0];

				if (btrfs_header_nritems(leaf) == 0) {
					WARN_ON(1);
					break;
				}

				/*
				 * Save the key so we can advances forward
				 * in the next search.
				 */
				btrfs_item_key_to_cpu(leaf, &key, 0);
				btrfs_release_path(path);
				root->cache_progress = last;
				mutex_unlock(&root->fs_commit_mutex);
				schedule_timeout(1);
				goto again;
			} else
				continue;
		}

		btrfs_item_key_to_cpu(leaf, &key, slot);

		if (key.type != BTRFS_INODE_ITEM_KEY)
			goto next;

		if (key.objectid >= root->highest_objectid)
			break;

		if (last != (u64)-1 && last + 1 != key.objectid) {
			__btrfs_add_free_space(ctl, last + 1,
					       key.objectid - last - 1);
			wake_up(&root->cache_wait);
		}

		last = key.objectid;
next:
		path->slots[0]++;
	}

	if (last < root->highest_objectid - 1) {
		__btrfs_add_free_space(ctl, last + 1,
				       root->highest_objectid - last - 1);
	}

	spin_lock(&root->cache_lock);
	root->cached = BTRFS_CACHE_FINISHED;
	spin_unlock(&root->cache_lock);

	root->cache_progress = (u64)-1;
	btrfs_unpin_free_ino(root);
out:
	wake_up(&root->cache_wait);
	mutex_unlock(&root->fs_commit_mutex);

	btrfs_free_path(path);

	return ret;
}

static void start_caching(struct btrfs_root *root)
{
	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
	struct task_struct *tsk;
	int ret;
	u64 objectid;

	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
		return;

	spin_lock(&root->cache_lock);
	if (root->cached != BTRFS_CACHE_NO) {
		spin_unlock(&root->cache_lock);
		return;
	}

	root->cached = BTRFS_CACHE_STARTED;
	spin_unlock(&root->cache_lock);

	ret = load_free_ino_cache(root->fs_info, root);
	if (ret == 1) {
		spin_lock(&root->cache_lock);
		root->cached = BTRFS_CACHE_FINISHED;
		spin_unlock(&root->cache_lock);
		return;
	}

	/*
	 * It can be quite time-consuming to fill the cache by searching
	 * through the extent tree, and this can keep ino allocation path
	 * waiting. Therefore at start we quickly find out the highest
	 * inode number and we know we can use inode numbers which fall in
	 * [highest_ino + 1, BTRFS_LAST_FREE_OBJECTID].
	 */
	ret = btrfs_find_free_objectid(root, &objectid);
	if (!ret && objectid <= BTRFS_LAST_FREE_OBJECTID) {
		__btrfs_add_free_space(ctl, objectid,
				       BTRFS_LAST_FREE_OBJECTID - objectid + 1);
	}

	tsk = kthread_run(caching_kthread, root, "btrfs-ino-cache-%llu\n",
			  root->root_key.objectid);
	BUG_ON(IS_ERR(tsk));
}

int btrfs_find_free_ino(struct btrfs_root *root, u64 *objectid)
{
	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
		return btrfs_find_free_objectid(root, objectid);

again:
	*objectid = btrfs_find_ino_for_alloc(root);

	if (*objectid != 0)
		return 0;

	start_caching(root);

	wait_event(root->cache_wait,
		   root->cached == BTRFS_CACHE_FINISHED ||
		   root->free_ino_ctl->free_space > 0);

	if (root->cached == BTRFS_CACHE_FINISHED &&
	    root->free_ino_ctl->free_space == 0)
		return -ENOSPC;
	else
		goto again;
}

void btrfs_return_ino(struct btrfs_root *root, u64 objectid)
{
	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
	struct btrfs_free_space_ctl *pinned = root->free_ino_pinned;

	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
		return;

again:
	if (root->cached == BTRFS_CACHE_FINISHED) {
		__btrfs_add_free_space(ctl, objectid, 1);
	} else {
		/*
		 * If we are in the process of caching free ino chunks,
		 * to avoid adding the same inode number to the free_ino
		 * tree twice due to cross transaction, we'll leave it
		 * in the pinned tree until a transaction is committed
		 * or the caching work is done.
		 */

		mutex_lock(&root->fs_commit_mutex);
		spin_lock(&root->cache_lock);
		if (root->cached == BTRFS_CACHE_FINISHED) {
			spin_unlock(&root->cache_lock);
			mutex_unlock(&root->fs_commit_mutex);
			goto again;
		}
		spin_unlock(&root->cache_lock);

		start_caching(root);

		if (objectid <= root->cache_progress ||
		    objectid > root->highest_objectid)
			__btrfs_add_free_space(ctl, objectid, 1);
		else
			__btrfs_add_free_space(pinned, objectid, 1);

		mutex_unlock(&root->fs_commit_mutex);
	}
}

/*
 * When a transaction is committed, we'll move those inode numbers which
 * are smaller than root->cache_progress from pinned tree to free_ino tree,
 * and others will just be dropped, because the commit root we were
 * searching has changed.
 *
 * Must be called with root->fs_commit_mutex held
 */
void btrfs_unpin_free_ino(struct btrfs_root *root)
{
	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
	struct rb_root *rbroot = &root->free_ino_pinned->free_space_offset;
	struct btrfs_free_space *info;
	struct rb_node *n;
	u64 count;

	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
		return;

	while (1) {
		n = rb_first(rbroot);
		if (!n)
			break;

		info = rb_entry(n, struct btrfs_free_space, offset_index);
		BUG_ON(info->bitmap);

		if (info->offset > root->cache_progress)
			goto free;
		else if (info->offset + info->bytes > root->cache_progress)
			count = root->cache_progress - info->offset + 1;
		else
			count = info->bytes;

		__btrfs_add_free_space(ctl, info->offset, count);
free:
		rb_erase(&info->offset_index, rbroot);
		kfree(info);
	}
}

#define INIT_THRESHOLD	(((1024 * 32) / 2) / sizeof(struct btrfs_free_space))
#define INODES_PER_BITMAP (PAGE_CACHE_SIZE * 8)

/*
 * The goal is to keep the memory used by the free_ino tree won't
 * exceed the memory if we use bitmaps only.
 */
static void recalculate_thresholds(struct btrfs_free_space_ctl *ctl)
{
	struct btrfs_free_space *info;
	struct rb_node *n;
	int max_ino;
	int max_bitmaps;

	n = rb_last(&ctl->free_space_offset);
	if (!n) {
		ctl->extents_thresh = INIT_THRESHOLD;
		return;
	}
	info = rb_entry(n, struct btrfs_free_space, offset_index);

	/*
	 * Find the maximum inode number in the filesystem. Note we
	 * ignore the fact that this can be a bitmap, because we are
	 * not doing precise calculation.
	 */
	max_ino = info->bytes - 1;

	max_bitmaps = ALIGN(max_ino, INODES_PER_BITMAP) / INODES_PER_BITMAP;
	if (max_bitmaps <= ctl->total_bitmaps) {
		ctl->extents_thresh = 0;
		return;
	}

	ctl->extents_thresh = (max_bitmaps - ctl->total_bitmaps) *
				PAGE_CACHE_SIZE / sizeof(*info);
}

/*
 * We don't fall back to bitmap, if we are below the extents threshold
 * or this chunk of inode numbers is a big one.
 */
static bool use_bitmap(struct btrfs_free_space_ctl *ctl,
		       struct btrfs_free_space *info)
{
	if (ctl->free_extents < ctl->extents_thresh ||
	    info->bytes > INODES_PER_BITMAP / 10)
		return false;

	return true;
}

static struct btrfs_free_space_op free_ino_op = {
	.recalc_thresholds	= recalculate_thresholds,
	.use_bitmap		= use_bitmap,
};

static void pinned_recalc_thresholds(struct btrfs_free_space_ctl *ctl)
{
}

static bool pinned_use_bitmap(struct btrfs_free_space_ctl *ctl,
			      struct btrfs_free_space *info)
{
	/*
	 * We always use extents for two reasons:
	 *
	 * - The pinned tree is only used during the process of caching
	 *   work.
	 * - Make code simpler. See btrfs_unpin_free_ino().
	 */
	return false;
}

static struct btrfs_free_space_op pinned_free_ino_op = {
	.recalc_thresholds	= pinned_recalc_thresholds,
	.use_bitmap		= pinned_use_bitmap,
};

void btrfs_init_free_ino_ctl(struct btrfs_root *root)
{
	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
	struct btrfs_free_space_ctl *pinned = root->free_ino_pinned;

	spin_lock_init(&ctl->tree_lock);
	ctl->unit = 1;
	ctl->start = 0;
	ctl->private = NULL;
	ctl->op = &free_ino_op;

	/*
	 * Initially we allow to use 16K of ram to cache chunks of
	 * inode numbers before we resort to bitmaps. This is somewhat
	 * arbitrary, but it will be adjusted in runtime.
	 */
	ctl->extents_thresh = INIT_THRESHOLD;

	spin_lock_init(&pinned->tree_lock);
	pinned->unit = 1;
	pinned->start = 0;
	pinned->private = NULL;
	pinned->extents_thresh = 0;
	pinned->op = &pinned_free_ino_op;
}

int btrfs_save_ino_cache(struct btrfs_root *root,
			 struct btrfs_trans_handle *trans)
{
	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
	struct btrfs_path *path;
	struct inode *inode;
	struct btrfs_block_rsv *rsv;
	u64 num_bytes;
	u64 alloc_hint = 0;
	int ret;
	int prealloc;
	bool retry = false;

	/* only fs tree and subvol/snap needs ino cache */
	if (root->root_key.objectid != BTRFS_FS_TREE_OBJECTID &&
	    (root->root_key.objectid < BTRFS_FIRST_FREE_OBJECTID ||
	     root->root_key.objectid > BTRFS_LAST_FREE_OBJECTID))
		return 0;

	/* Don't save inode cache if we are deleting this root */
	if (btrfs_root_refs(&root->root_item) == 0 &&
	    root != root->fs_info->tree_root)
		return 0;

	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
		return 0;

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;

	rsv = trans->block_rsv;
	trans->block_rsv = &root->fs_info->trans_block_rsv;

	num_bytes = trans->bytes_reserved;
	/*
	 * 1 item for inode item insertion if need
	 * 3 items for inode item update (in the worst case)
	 * 1 item for free space object
	 * 3 items for pre-allocation
	 */
	trans->bytes_reserved = btrfs_calc_trans_metadata_size(root, 8);
	ret = btrfs_block_rsv_add_noflush(root, trans->block_rsv,
					  trans->bytes_reserved);
	if (ret)
		goto out;
	trace_btrfs_space_reservation(root->fs_info, "ino_cache",
				      (u64)(unsigned long)trans,
				      trans->bytes_reserved, 1);
again:
	inode = lookup_free_ino_inode(root, path);
	if (IS_ERR(inode) && PTR_ERR(inode) != -ENOENT) {
		ret = PTR_ERR(inode);
		goto out_release;
	}

	if (IS_ERR(inode)) {
		BUG_ON(retry);
		retry = true;

		ret = create_free_ino_inode(root, trans, path);
		if (ret)
			goto out_release;
		goto again;
	}

	BTRFS_I(inode)->generation = 0;
	ret = btrfs_update_inode(trans, root, inode);
	WARN_ON(ret);

	if (i_size_read(inode) > 0) {
		ret = btrfs_truncate_free_space_cache(root, trans, path, inode);
		if (ret)
			goto out_put;
	}

	spin_lock(&root->cache_lock);
	if (root->cached != BTRFS_CACHE_FINISHED) {
		ret = -1;
		spin_unlock(&root->cache_lock);
		goto out_put;
	}
	spin_unlock(&root->cache_lock);

	spin_lock(&ctl->tree_lock);
	prealloc = sizeof(struct btrfs_free_space) * ctl->free_extents;
	prealloc = ALIGN(prealloc, PAGE_CACHE_SIZE);
	prealloc += ctl->total_bitmaps * PAGE_CACHE_SIZE;
	spin_unlock(&ctl->tree_lock);

	/* Just to make sure we have enough space */
	prealloc += 8 * PAGE_CACHE_SIZE;

	ret = btrfs_delalloc_reserve_space(inode, prealloc);
	if (ret)
		goto out_put;

	ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, prealloc,
					      prealloc, prealloc, &alloc_hint);
	if (ret) {
		btrfs_delalloc_release_space(inode, prealloc);
		goto out_put;
	}
	btrfs_free_reserved_data_space(inode, prealloc);

	ret = btrfs_write_out_ino_cache(root, trans, path);
out_put:
	iput(inode);
out_release:
	trace_btrfs_space_reservation(root->fs_info, "ino_cache",
				      (u64)(unsigned long)trans,
				      trans->bytes_reserved, 0);
	btrfs_block_rsv_release(root, trans->block_rsv, trans->bytes_reserved);
out:
	trans->block_rsv = rsv;
	trans->bytes_reserved = num_bytes;

	btrfs_free_path(path);
	return ret;
}

static int btrfs_find_highest_objectid(struct btrfs_root *root, u64 *objectid)
{
	struct btrfs_path *path;
	int ret;
	struct extent_buffer *l;
	struct btrfs_key search_key;
	struct btrfs_key found_key;
	int slot;

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;

	search_key.objectid = BTRFS_LAST_FREE_OBJECTID;
	search_key.type = -1;
	search_key.offset = (u64)-1;
	ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
	if (ret < 0)
		goto error;
	BUG_ON(ret == 0);
	if (path->slots[0] > 0) {
		slot = path->slots[0] - 1;
		l = path->nodes[0];
		btrfs_item_key_to_cpu(l, &found_key, slot);
		*objectid = max_t(u64, found_key.objectid,
				  BTRFS_FIRST_FREE_OBJECTID - 1);
	} else {
		*objectid = BTRFS_FIRST_FREE_OBJECTID - 1;
	}
	ret = 0;
error:
	btrfs_free_path(path);
	return ret;
}

int btrfs_find_free_objectid(struct btrfs_root *root, u64 *objectid)
{
	int ret;
	mutex_lock(&root->objectid_mutex);

	if (unlikely(root->highest_objectid < BTRFS_FIRST_FREE_OBJECTID)) {
		ret = btrfs_find_highest_objectid(root,
						  &root->highest_objectid);
		if (ret)
			goto out;
	}

	if (unlikely(root->highest_objectid >= BTRFS_LAST_FREE_OBJECTID)) {
		ret = -ENOSPC;
		goto out;
	}

	*objectid = ++root->highest_objectid;
	ret = 0;
out:
	mutex_unlock(&root->objectid_mutex);
	return ret;
}