summaryrefslogtreecommitdiffstats
path: root/fs/ubifs/budget.c
blob: 44cff803171a9cf12fc4e4a8339b92b728dc2060 (plain)
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
/*
 * This file is part of UBIFS.
 *
 * Copyright (C) 2006-2008 Nokia Corporation.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License version 2 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., 51
 * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
 *
 * Authors: Adrian Hunter
 *          Artem Bityutskiy (Битюцкий Артём)
 */

/*
 * This file implements the budgeting sub-system which is responsible for UBIFS
 * space management.
 *
 * Factors such as compression, wasted space at the ends of LEBs, space in other
 * journal heads, the effect of updates on the index, and so on, make it
 * impossible to accurately predict the amount of space needed. Consequently
 * approximations are used.
 */

#include "ubifs.h"
#include <linux/writeback.h>
#include <linux/math64.h>

/*
 * When pessimistic budget calculations say that there is no enough space,
 * UBIFS starts writing back dirty inodes and pages, doing garbage collection,
 * or committing. The below constants define maximum number of times UBIFS
 * repeats the operations.
 */
#define MAX_SHRINK_RETRIES 8
#define MAX_GC_RETRIES     4
#define MAX_CMT_RETRIES    2
#define MAX_NOSPC_RETRIES  1

/*
 * The below constant defines amount of dirty pages which should be written
 * back at when trying to shrink the liability.
 */
#define NR_TO_WRITE 16

/**
 * struct retries_info - information about re-tries while making free space.
 * @prev_liability: previous liability
 * @shrink_cnt: how many times the liability was shrinked
 * @shrink_retries: count of liability shrink re-tries (increased when
 *                  liability does not shrink)
 * @try_gc: GC should be tried first
 * @gc_retries: how many times GC was run
 * @cmt_retries: how many times commit has been done
 * @nospc_retries: how many times GC returned %-ENOSPC
 *
 * Since we consider budgeting to be the fast-path, and this structure has to
 * be allocated on stack and zeroed out, we make it smaller using bit-fields.
 */
struct retries_info {
	long long prev_liability;
	unsigned int shrink_cnt;
	unsigned int shrink_retries:5;
	unsigned int try_gc:1;
	unsigned int gc_retries:4;
	unsigned int cmt_retries:3;
	unsigned int nospc_retries:1;
};

/**
 * shrink_liability - write-back some dirty pages/inodes.
 * @c: UBIFS file-system description object
 * @nr_to_write: how many dirty pages to write-back
 *
 * This function shrinks UBIFS liability by means of writing back some amount
 * of dirty inodes and their pages. Returns the amount of pages which were
 * written back. The returned value does not include dirty inodes which were
 * synchronized.
 *
 * Note, this function synchronizes even VFS inodes which are locked
 * (@i_mutex) by the caller of the budgeting function, because write-back does
 * not touch @i_mutex.
 */
static int shrink_liability(struct ubifs_info *c, int nr_to_write)
{
	int nr_written;
	struct writeback_control wbc = {
		.sync_mode   = WB_SYNC_NONE,
		.range_end   = LLONG_MAX,
		.nr_to_write = nr_to_write,
	};

	generic_sync_sb_inodes(c->vfs_sb, &wbc);
	nr_written = nr_to_write - wbc.nr_to_write;

	if (!nr_written) {
		/*
		 * Re-try again but wait on pages/inodes which are being
		 * written-back concurrently (e.g., by pdflush).
		 */
		memset(&wbc, 0, sizeof(struct writeback_control));
		wbc.sync_mode   = WB_SYNC_ALL;
		wbc.range_end   = LLONG_MAX;
		wbc.nr_to_write = nr_to_write;
		generic_sync_sb_inodes(c->vfs_sb, &wbc);
		nr_written = nr_to_write - wbc.nr_to_write;
	}

	dbg_budg("%d pages were written back", nr_written);
	return nr_written;
}


/**
 * run_gc - run garbage collector.
 * @c: UBIFS file-system description object
 *
 * This function runs garbage collector to make some more free space. Returns
 * zero if a free LEB has been produced, %-EAGAIN if commit is required, and a
 * negative error code in case of failure.
 */
static int run_gc(struct ubifs_info *c)
{
	int err, lnum;

	/* Make some free space by garbage-collecting dirty space */
	down_read(&c->commit_sem);
	lnum = ubifs_garbage_collect(c, 1);
	up_read(&c->commit_sem);
	if (lnum < 0)
		return lnum;

	/* GC freed one LEB, return it to lprops */
	dbg_budg("GC freed LEB %d", lnum);
	err = ubifs_return_leb(c, lnum);
	if (err)
		return err;
	return 0;
}

/**
 * make_free_space - make more free space on the file-system.
 * @c: UBIFS file-system description object
 * @ri: information about previous invocations of this function
 *
 * This function is called when an operation cannot be budgeted because there
 * is supposedly no free space. But in most cases there is some free space:
 *   o budgeting is pessimistic, so it always budgets more then it is actually
 *     needed, so shrinking the liability is one way to make free space - the
 *     cached data will take less space then it was budgeted for;
 *   o GC may turn some dark space into free space (budgeting treats dark space
 *     as not available);
 *   o commit may free some LEB, i.e., turn freeable LEBs into free LEBs.
 *
 * So this function tries to do the above. Returns %-EAGAIN if some free space
 * was presumably made and the caller has to re-try budgeting the operation.
 * Returns %-ENOSPC if it couldn't do more free space, and other negative error
 * codes on failures.
 */
static int make_free_space(struct ubifs_info *c, struct retries_info *ri)
{
	int err;

	/*
	 * If we have some dirty pages and inodes (liability), try to write
	 * them back unless this was tried too many times without effect
	 * already.
	 */
	if (ri->shrink_retries < MAX_SHRINK_RETRIES && !ri->try_gc) {
		long long liability;

		spin_lock(&c->space_lock);
		liability = c->budg_idx_growth + c->budg_data_growth +
			    c->budg_dd_growth;
		spin_unlock(&c->space_lock);

		if (ri->prev_liability >= liability) {
			/* Liability does not shrink, next time try GC then */
			ri->shrink_retries += 1;
			if (ri->gc_retries < MAX_GC_RETRIES)
				ri->try_gc = 1;
			dbg_budg("liability did not shrink: retries %d of %d",
				 ri->shrink_retries, MAX_SHRINK_RETRIES);
		}

		dbg_budg("force write-back (count %d)", ri->shrink_cnt);
		shrink_liability(c, NR_TO_WRITE + ri->shrink_cnt);

		ri->prev_liability = liability;
		ri->shrink_cnt += 1;
		return -EAGAIN;
	}

	/*
	 * Try to run garbage collector unless it was already tried too many
	 * times.
	 */
	if (ri->gc_retries < MAX_GC_RETRIES) {
		ri->gc_retries += 1;
		dbg_budg("run GC, retries %d of %d",
			 ri->gc_retries, MAX_GC_RETRIES);

		ri->try_gc = 0;
		err = run_gc(c);
		if (!err)
			return -EAGAIN;

		if (err == -EAGAIN) {
			dbg_budg("GC asked to commit");
			err = ubifs_run_commit(c);
			if (err)
				return err;
			return -EAGAIN;
		}

		if (err != -ENOSPC)
			return err;

		/*
		 * GC could not make any progress. If this is the first time,
		 * then it makes sense to try to commit, because it might make
		 * some dirty space.
		 */
		dbg_budg("GC returned -ENOSPC, retries %d",
			 ri->nospc_retries);
		if (ri->nospc_retries >= MAX_NOSPC_RETRIES)
			return err;
		ri->nospc_retries += 1;
	}

	/* Neither GC nor write-back helped, try to commit */
	if (ri->cmt_retries < MAX_CMT_RETRIES) {
		ri->cmt_retries += 1;
		dbg_budg("run commit, retries %d of %d",
			 ri->cmt_retries, MAX_CMT_RETRIES);
		err = ubifs_run_commit(c);
		if (err)
			return err;
		return -EAGAIN;
	}
	return -ENOSPC;
}

/**
 * ubifs_calc_min_idx_lebs - calculate amount of eraseblocks for the index.
 * @c: UBIFS file-system description object
 *
 * This function calculates and returns the number of eraseblocks which should
 * be kept for index usage.
 */
int ubifs_calc_min_idx_lebs(struct ubifs_info *c)
{
	int idx_lebs, eff_leb_size = c->leb_size - c->max_idx_node_sz;
	long long idx_size;

	idx_size = c->old_idx_sz + c->budg_idx_growth + c->budg_uncommitted_idx;

	/* And make sure we have thrice the index size of space reserved */
	idx_size = idx_size + (idx_size << 1);

	/*
	 * We do not maintain 'old_idx_size' as 'old_idx_lebs'/'old_idx_bytes'
	 * pair, nor similarly the two variables for the new index size, so we
	 * have to do this costly 64-bit division on fast-path.
	 */
	idx_size += eff_leb_size - 1;
	idx_lebs = div_u64(idx_size, eff_leb_size);
	/*
	 * The index head is not available for the in-the-gaps method, so add an
	 * extra LEB to compensate.
	 */
	idx_lebs += 1;
	if (idx_lebs < MIN_INDEX_LEBS)
		idx_lebs = MIN_INDEX_LEBS;
	return idx_lebs;
}

/**
 * ubifs_calc_available - calculate available FS space.
 * @c: UBIFS file-system description object
 * @min_idx_lebs: minimum number of LEBs reserved for the index
 *
 * This function calculates and returns amount of FS space available for use.
 */
long long ubifs_calc_available(const struct ubifs_info *c, int min_idx_lebs)
{
	int subtract_lebs;
	long long available;

	available = c->main_bytes - c->lst.total_used;

	/*
	 * Now 'available' contains theoretically available flash space
	 * assuming there is no index, so we have to subtract the space which
	 * is reserved for the index.
	 */
	subtract_lebs = min_idx_lebs;

	/* Take into account that GC reserves one LEB for its own needs */
	subtract_lebs += 1;

	/*
	 * The GC journal head LEB is not really accessible. And since
	 * different write types go to different heads, we may count only on
	 * one head's space.
	 */
	subtract_lebs += c->jhead_cnt - 1;

	/* We also reserve one LEB for deletions, which bypass budgeting */
	subtract_lebs += 1;

	available -= (long long)subtract_lebs * c->leb_size;

	/* Subtract the dead space which is not available for use */
	available -= c->lst.total_dead;

	/*
	 * Subtract dark space, which might or might not be usable - it depends
	 * on the data which we have on the media and which will be written. If
	 * this is a lot of uncompressed or not-compressible data, the dark
	 * space cannot be used.
	 */
	available -= c->lst.total_dark;

	/*
	 * However, there is more dark space. The index may be bigger than
	 * @min_idx_lebs. Those extra LEBs are assumed to be available, but
	 * their dark space is not included in total_dark, so it is subtracted
	 * here.
	 */
	if (c->lst.idx_lebs > min_idx_lebs) {
		subtract_lebs = c->lst.idx_lebs - min_idx_lebs;
		available -= subtract_lebs * c->dark_wm;
	}

	/* The calculations are rough and may end up with a negative number */
	return available > 0 ? available : 0;
}

/**
 * can_use_rp - check whether the user is allowed to use reserved pool.
 * @c: UBIFS file-system description object
 *
 * UBIFS has so-called "reserved pool" which is flash space reserved
 * for the superuser and for uses whose UID/GID is recorded in UBIFS superblock.
 * This function checks whether current user is allowed to use reserved pool.
 * Returns %1  current user is allowed to use reserved pool and %0 otherwise.
 */
static int can_use_rp(struct ubifs_info *c)
{
	if (current->fsuid == c->rp_uid || capable(CAP_SYS_RESOURCE) ||
	    (c->rp_gid != 0 && in_group_p(c->rp_gid)))
		return 1;
	return 0;
}

/**
 * do_budget_space - reserve flash space for index and data growth.
 * @c: UBIFS file-system description object
 *
 * This function makes sure UBIFS has enough free eraseblocks for index growth
 * and data.
 *
 * When budgeting index space, UBIFS reserves thrice as many LEBs as the index
 * would take if it was consolidated and written to the flash. This guarantees
 * that the "in-the-gaps" commit method always succeeds and UBIFS will always
 * be able to commit dirty index. So this function basically adds amount of
 * budgeted index space to the size of the current index, multiplies this by 3,
 * and makes sure this does not exceed the amount of free eraseblocks.
 *
 * Notes about @c->min_idx_lebs and @c->lst.idx_lebs variables:
 * o @c->lst.idx_lebs is the number of LEBs the index currently uses. It might
 *    be large, because UBIFS does not do any index consolidation as long as
 *    there is free space. IOW, the index may take a lot of LEBs, but the LEBs
 *    will contain a lot of dirt.
 * o @c->min_idx_lebs is the the index presumably takes. IOW, the index may be
 *   consolidated to take up to @c->min_idx_lebs LEBs.
 *
 * This function returns zero in case of success, and %-ENOSPC in case of
 * failure.
 */
static int do_budget_space(struct ubifs_info *c)
{
	long long outstanding, available;
	int lebs, rsvd_idx_lebs, min_idx_lebs;

	/* First budget index space */
	min_idx_lebs = ubifs_calc_min_idx_lebs(c);

	/* Now 'min_idx_lebs' contains number of LEBs to reserve */
	if (min_idx_lebs > c->lst.idx_lebs)
		rsvd_idx_lebs = min_idx_lebs - c->lst.idx_lebs;
	else
		rsvd_idx_lebs = 0;

	/*
	 * The number of LEBs that are available to be used by the index is:
	 *
	 *    @c->lst.empty_lebs + @c->freeable_cnt + @c->idx_gc_cnt -
	 *    @c->lst.taken_empty_lebs
	 *
	 * @c->lst.empty_lebs are available because they are empty.
	 * @c->freeable_cnt are available because they contain only free and
	 * dirty space, @c->idx_gc_cnt are available because they are index
	 * LEBs that have been garbage collected and are awaiting the commit
	 * before they can be used. And the in-the-gaps method will grab these
	 * if it needs them. @c->lst.taken_empty_lebs are empty LEBs that have
	 * already been allocated for some purpose.
	 *
	 * Note, @c->idx_gc_cnt is included to both @c->lst.empty_lebs (because
	 * these LEBs are empty) and to @c->lst.taken_empty_lebs (because they
	 * are taken until after the commit).
	 *
	 * Note, @c->lst.taken_empty_lebs may temporarily be higher by one
	 * because of the way we serialize LEB allocations and budgeting. See a
	 * comment in 'ubifs_find_free_space()'.
	 */
	lebs = c->lst.empty_lebs + c->freeable_cnt + c->idx_gc_cnt -
	       c->lst.taken_empty_lebs;
	if (unlikely(rsvd_idx_lebs > lebs)) {
		dbg_budg("out of indexing space: min_idx_lebs %d (old %d), "
			 "rsvd_idx_lebs %d", min_idx_lebs, c->min_idx_lebs,
			 rsvd_idx_lebs);
		return -ENOSPC;
	}

	available = ubifs_calc_available(c, min_idx_lebs);
	outstanding = c->budg_data_growth + c->budg_dd_growth;

	if (unlikely(available < outstanding)) {
		dbg_budg("out of data space: available %lld, outstanding %lld",
			 available, outstanding);
		return -ENOSPC;
	}

	if (available - outstanding <= c->rp_size && !can_use_rp(c))
		return -ENOSPC;

	c->min_idx_lebs = min_idx_lebs;
	return 0;
}

/**
 * calc_idx_growth - calculate approximate index growth from budgeting request.
 * @c: UBIFS file-system description object
 * @req: budgeting request
 *
 * For now we assume each new node adds one znode. But this is rather poor
 * approximation, though.
 */
static int calc_idx_growth(const struct ubifs_info *c,
			   const struct ubifs_budget_req *req)
{
	int znodes;

	znodes = req->new_ino + (req->new_page << UBIFS_BLOCKS_PER_PAGE_SHIFT) +
		 req->new_dent;
	return znodes * c->max_idx_node_sz;
}

/**
 * calc_data_growth - calculate approximate amount of new data from budgeting
 * request.
 * @c: UBIFS file-system description object
 * @req: budgeting request
 */
static int calc_data_growth(const struct ubifs_info *c,
			    const struct ubifs_budget_req *req)
{
	int data_growth;

	data_growth = req->new_ino  ? c->inode_budget : 0;
	if (req->new_page)
		data_growth += c->page_budget;
	if (req->new_dent)
		data_growth += c->dent_budget;
	data_growth += req->new_ino_d;
	return data_growth;
}

/**
 * calc_dd_growth - calculate approximate amount of data which makes other data
 * dirty from budgeting request.
 * @c: UBIFS file-system description object
 * @req: budgeting request
 */
static int calc_dd_growth(const struct ubifs_info *c,
			  const struct ubifs_budget_req *req)
{
	int dd_growth;

	dd_growth = req->dirtied_page ? c->page_budget : 0;

	if (req->dirtied_ino)
		dd_growth += c->inode_budget << (req->dirtied_ino - 1);
	if (req->mod_dent)
		dd_growth += c->dent_budget;
	dd_growth += req->dirtied_ino_d;
	return dd_growth;
}

/**
 * ubifs_budget_space - ensure there is enough space to complete an operation.
 * @c: UBIFS file-system description object
 * @req: budget request
 *
 * This function allocates budget for an operation. It uses pessimistic
 * approximation of how much flash space the operation needs. The goal of this
 * function is to make sure UBIFS always has flash space to flush all dirty
 * pages, dirty inodes, and dirty znodes (liability). This function may force
 * commit, garbage-collection or write-back. Returns zero in case of success,
 * %-ENOSPC if there is no free space and other negative error codes in case of
 * failures.
 */
int ubifs_budget_space(struct ubifs_info *c, struct ubifs_budget_req *req)
{
	int uninitialized_var(cmt_retries), uninitialized_var(wb_retries);
	int err, idx_growth, data_growth, dd_growth;
	struct retries_info ri;

	ubifs_assert(req->new_page <= 1);
	ubifs_assert(req->dirtied_page <= 1);
	ubifs_assert(req->new_dent <= 1);
	ubifs_assert(req->mod_dent <= 1);
	ubifs_assert(req->new_ino <= 1);
	ubifs_assert(req->new_ino_d <= UBIFS_MAX_INO_DATA);
	ubifs_assert(req->dirtied_ino <= 4);
	ubifs_assert(req->dirtied_ino_d <= UBIFS_MAX_INO_DATA * 4);
	ubifs_assert(!(req->new_ino_d & 7));
	ubifs_assert(!(req->dirtied_ino_d & 7));

	data_growth = calc_data_growth(c, req);
	dd_growth = calc_dd_growth(c, req);
	if (!data_growth && !dd_growth)
		return 0;
	idx_growth = calc_idx_growth(c, req);
	memset(&ri, 0, sizeof(struct retries_info));

again:
	spin_lock(&c->space_lock);
	ubifs_assert(c->budg_idx_growth >= 0);
	ubifs_assert(c->budg_data_growth >= 0);
	ubifs_assert(c->budg_dd_growth >= 0);

	if (unlikely(c->nospace) && (c->nospace_rp || !can_use_rp(c))) {
		dbg_budg("no space");
		spin_unlock(&c->space_lock);
		return -ENOSPC;
	}

	c->budg_idx_growth += idx_growth;
	c->budg_data_growth += data_growth;
	c->budg_dd_growth += dd_growth;

	err = do_budget_space(c);
	if (likely(!err)) {
		req->idx_growth = idx_growth;
		req->data_growth = data_growth;
		req->dd_growth = dd_growth;
		spin_unlock(&c->space_lock);
		return 0;
	}

	/* Restore the old values */
	c->budg_idx_growth -= idx_growth;
	c->budg_data_growth -= data_growth;
	c->budg_dd_growth -= dd_growth;
	spin_unlock(&c->space_lock);

	if (req->fast) {
		dbg_budg("no space for fast budgeting");
		return err;
	}

	err = make_free_space(c, &ri);
	if (err == -EAGAIN) {
		dbg_budg("try again");
		cond_resched();
		goto again;
	} else if (err == -ENOSPC) {
		dbg_budg("FS is full, -ENOSPC");
		c->nospace = 1;
		if (can_use_rp(c) || c->rp_size == 0)
			c->nospace_rp = 1;
		smp_wmb();
	} else
		ubifs_err("cannot budget space, error %d", err);
	return err;
}

/**
 * ubifs_release_budget - release budgeted free space.
 * @c: UBIFS file-system description object
 * @req: budget request
 *
 * This function releases the space budgeted by 'ubifs_budget_space()'. Note,
 * since the index changes (which were budgeted for in @req->idx_growth) will
 * only be written to the media on commit, this function moves the index budget
 * from @c->budg_idx_growth to @c->budg_uncommitted_idx. The latter will be
 * zeroed by the commit operation.
 */
void ubifs_release_budget(struct ubifs_info *c, struct ubifs_budget_req *req)
{
	ubifs_assert(req->new_page <= 1);
	ubifs_assert(req->dirtied_page <= 1);
	ubifs_assert(req->new_dent <= 1);
	ubifs_assert(req->mod_dent <= 1);
	ubifs_assert(req->new_ino <= 1);
	ubifs_assert(req->new_ino_d <= UBIFS_MAX_INO_DATA);
	ubifs_assert(req->dirtied_ino <= 4);
	ubifs_assert(req->dirtied_ino_d <= UBIFS_MAX_INO_DATA * 4);
	ubifs_assert(!(req->new_ino_d & 7));
	ubifs_assert(!(req->dirtied_ino_d & 7));
	if (!req->recalculate) {
		ubifs_assert(req->idx_growth >= 0);
		ubifs_assert(req->data_growth >= 0);
		ubifs_assert(req->dd_growth >= 0);
	}

	if (req->recalculate) {
		req->data_growth = calc_data_growth(c, req);
		req->dd_growth = calc_dd_growth(c, req);
		req->idx_growth = calc_idx_growth(c, req);
	}

	if (!req->data_growth && !req->dd_growth)
		return;

	c->nospace = c->nospace_rp = 0;
	smp_wmb();

	spin_lock(&c->space_lock);
	c->budg_idx_growth -= req->idx_growth;
	c->budg_uncommitted_idx += req->idx_growth;
	c->budg_data_growth -= req->data_growth;
	c->budg_dd_growth -= req->dd_growth;
	c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);

	ubifs_assert(c->budg_idx_growth >= 0);
	ubifs_assert(c->budg_data_growth >= 0);
	ubifs_assert(c->budg_dd_growth >= 0);
	ubifs_assert(c->min_idx_lebs < c->main_lebs);
	ubifs_assert(!(c->budg_idx_growth & 7));
	ubifs_assert(!(c->budg_data_growth & 7));
	ubifs_assert(!(c->budg_dd_growth & 7));
	spin_unlock(&c->space_lock);
}

/**
 * ubifs_convert_page_budget - convert budget of a new page.
 * @c: UBIFS file-system description object
 *
 * This function converts budget which was allocated for a new page of data to
 * the budget of changing an existing page of data. The latter is smaller then
 * the former, so this function only does simple re-calculation and does not
 * involve any write-back.
 */
void ubifs_convert_page_budget(struct ubifs_info *c)
{
	spin_lock(&c->space_lock);
	/* Release the index growth reservation */
	c->budg_idx_growth -= c->max_idx_node_sz << UBIFS_BLOCKS_PER_PAGE_SHIFT;
	/* Release the data growth reservation */
	c->budg_data_growth -= c->page_budget;
	/* Increase the dirty data growth reservation instead */
	c->budg_dd_growth += c->page_budget;
	/* And re-calculate the indexing space reservation */
	c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
	spin_unlock(&c->space_lock);
}

/**
 * ubifs_release_dirty_inode_budget - release dirty inode budget.
 * @c: UBIFS file-system description object
 * @ui: UBIFS inode to release the budget for
 *
 * This function releases budget corresponding to a dirty inode. It is usually
 * called when after the inode has been written to the media and marked as
 * clean.
 */
void ubifs_release_dirty_inode_budget(struct ubifs_info *c,
				      struct ubifs_inode *ui)
{
	struct ubifs_budget_req req;

	memset(&req, 0, sizeof(struct ubifs_budget_req));
	req.dd_growth = c->inode_budget + ALIGN(ui->data_len, 8);
	ubifs_release_budget(c, &req);
}

/**
 * ubifs_reported_space - calculate reported free space.
 * @c: the UBIFS file-system description object
 * @free: amount of free space
 *
 * This function calculates amount of free space which will be reported to
 * user-space. User-space application tend to expect that if the file-system
 * (e.g., via the 'statfs()' call) reports that it has N bytes available, they
 * are able to write a file of size N. UBIFS attaches node headers to each data
 * node and it has to write indexind nodes as well. This introduces additional
 * overhead, and UBIFS has to report sligtly less free space to meet the above
 * expectetions.
 *
 * This function assumes free space is made up of uncompressed data nodes and
 * full index nodes (one per data node, tripled because we always allow enough
 * space to write the index thrice).
 *
 * Note, the calculation is pessimistic, which means that most of the time
 * UBIFS reports less space than it actually has.
 */
long long ubifs_reported_space(const struct ubifs_info *c, long long free)
{
	int divisor, factor, f;

	/*
	 * Reported space size is @free * X, where X is UBIFS block size
	 * divided by UBIFS block size + all overhead one data block
	 * introduces. The overhead is the node header + indexing overhead.
	 *
	 * Indexing overhead calculations are based on the following formula:
	 * I = N/(f - 1) + 1, where I - number of indexing nodes, N - number
	 * of data nodes, f - fanout. Because effective UBIFS fanout is twice
	 * as less than maximum fanout, we assume that each data node
	 * introduces 3 * @c->max_idx_node_sz / (@c->fanout/2 - 1) bytes.
	 * Note, the multiplier 3 is because UBIFS reseves thrice as more space
	 * for the index.
	 */
	f = c->fanout > 3 ? c->fanout >> 1 : 2;
	factor = UBIFS_BLOCK_SIZE;
	divisor = UBIFS_MAX_DATA_NODE_SZ;
	divisor += (c->max_idx_node_sz * 3) / (f - 1);
	free *= factor;
	return div_u64(free, divisor);
}

/**
 * ubifs_get_free_space - return amount of free space.
 * @c: UBIFS file-system description object
 *
 * This function calculates amount of free space to report to user-space.
 *
 * Because UBIFS may introduce substantial overhead (the index, node headers,
 * alighment, wastage at the end of eraseblocks, etc), it cannot report real
 * amount of free flash space it has (well, because not all dirty space is
 * reclamable, UBIFS does not actually know the real amount). If UBIFS did so,
 * it would bread user expectetion about what free space is. Users seem to
 * accustomed to assume that if the file-system reports N bytes of free space,
 * they would be able to fit a file of N bytes to the FS. This almost works for
 * traditional file-systems, because they have way less overhead than UBIFS.
 * So, to keep users happy, UBIFS tries to take the overhead into account.
 */
long long ubifs_get_free_space(struct ubifs_info *c)
{
	int min_idx_lebs, rsvd_idx_lebs, lebs;
	long long available, outstanding, free;

	spin_lock(&c->space_lock);
	min_idx_lebs = c->min_idx_lebs;
	ubifs_assert(min_idx_lebs == ubifs_calc_min_idx_lebs(c));
	outstanding = c->budg_data_growth + c->budg_dd_growth;

	/*
	 * Force the amount available to the total size reported if the used
	 * space is zero.
	 */
	if (c->lst.total_used <= UBIFS_INO_NODE_SZ && !outstanding) {
		spin_unlock(&c->space_lock);
		return (long long)c->block_cnt << UBIFS_BLOCK_SHIFT;
	}

	available = ubifs_calc_available(c, min_idx_lebs);

	/*
	 * When reporting free space to user-space, UBIFS guarantees that it is
	 * possible to write a file of free space size. This means that for
	 * empty LEBs we may use more precise calculations than
	 * 'ubifs_calc_available()' is using. Namely, we know that in empty
	 * LEBs we would waste only @c->leb_overhead bytes, not @c->dark_wm.
	 * Thus, amend the available space.
	 *
	 * Note, the calculations below are similar to what we have in
	 * 'do_budget_space()', so refer there for comments.
	 */
	if (min_idx_lebs > c->lst.idx_lebs)
		rsvd_idx_lebs = min_idx_lebs - c->lst.idx_lebs;
	else
		rsvd_idx_lebs = 0;
	lebs = c->lst.empty_lebs + c->freeable_cnt + c->idx_gc_cnt -
	       c->lst.taken_empty_lebs;
	lebs -= rsvd_idx_lebs;
	available += lebs * (c->dark_wm - c->leb_overhead);
	spin_unlock(&c->space_lock);

	if (available > outstanding)
		free = ubifs_reported_space(c, available - outstanding);
	else
		free = 0;
	return free;
}
OpenPOWER on IntegriCloud